Partially presented at the Annual Meeting of the American Society of Clinical Oncology, Orlando, Florida, May 29-June 2, 2009.
Refusal of appropriately indicated do-not-resuscitate (DNR) orders may cause harm and distress for patients, families, and the medical team. We conducted a retrospective study to determine the frequency and predictors of refusals of DNR in advanced cancer patients admitted to an acute palliative care unit.
A total of 2538 consecutive admissions were reviewed. Demographic and clinical characteristics from 200 consecutive patients with DNR orders and 100 consecutive patients who refused DNR were collected, and differences between the groups were determined by multivariate regression and recursive partitioning analysis.
Of 2538 admissions, 2530 (99%) were appropriate for DNR discussion. Of the 2530 admissions, 2374 were unique patients, and 100 (4%) of 2374 refused DNR. Refusers had median (interquartile range, IQR) pain of 7 (4-9) versus 5 (3-8, P = .0005), nausea of 2 (0-7) versus 1 (0-4, P = .05), and dyspnea of 1 (0-5) versus 4 (0-7, P = .002) as compared with DNR nonrefusers, respectively. Patients with hematological malignancies and advance directives had a lower DNR refusal risk (odds ratio [OR], 0.38; P = .02, and OR, 0.36; P < .0001, respectively). Multivariate regression analysis revealed that patients with moderate-severe pain (OR, 3.19; P = .002) and with no advance directives (OR, 2.94; P ≤ .001) had higher DNR refusal risk. There were more inpatient deaths among DNR nonrefusers (87 of 200 vs 1 of 100, P < .0001). Median (IQR) time from discharge to death was 18 (8-35) days for those with DNR orders and 85 (25-206) days for DNR refusers (P ≤ .0001).
Cardiopulmonary resuscitation (CPR) was proposed in the early 1960s as an alternative to open thoracotomy and internal cardiac massage for cardiac arrest that occurred without an immediately apparent reason in the cardiac perioperative setting.1, 2 The initial reports on the efficacy of the procedure were very promising, reaching figures of around 95%.3 CPR thus evolved as the standard of care for all patients who die during hospital admission, although these studies were conducted in very specific populations, and therefore lacked generalizability. Currently, CPR is the default condition in the United States, and is an invasive procedure provided even in the absence of patient consent and if no contrary request is available. The do-not-resuscitate (DNR) order is the legal way to document that CPR is not to take place, allowing healthcare providers not to start CPR in case of cardiac arrest.4 Typically, DNR status is implemented when CPR is not medically indicated because of lack of efficacy and/or when consistent with the individual's expressed wishes. In these lines, there has been debate around the appropriateness of the term “do not resuscitate,” and the wording “allow natural death” has been proposed as preferable.5-7
In advanced cancer patients, it has been consistently shown that favorable outcomes of CPR are extremely rare.8 In this setting, CPR has been described as harmful for a variety of reasons, such as physical suffering, loss of dignity, prolonged death, survival with unacceptable quality of life, and increasing family suffering.9 DNR status is appropriate for the majority of advanced cancer patients, who have been shown to be willing to discuss end-of-life care.10, 11 Therefore, it has been proposed that end-of-life care and CPR should be discussed with all cancer patients at hospital admission, to provide information regarding realistic expectations and to allow the medical staff to document patients' preferences.12
Advance directives are documents that allow patients to exercise control over healthcare decisions in the event that they lose decision-making capacity.13 Because the majority of cancer patients die in delirium, and are therefore unable to make decisions,14, 15 advance directives are especially advisable in this population.
The purpose of this case-control study was to determine the frequency of refusal of appropriate DNR orders by advanced cancer patients admitted to the inpatient acute palliative care unit and to determine the characteristics of the patients who refused these orders. We also sought to determine the discharge clinical outcomes and resuscitation status of advanced cancer patients who refused DNR orders, and to determine whether the presence of a completed advance directive available in the medical record was associated with patient acceptance of medical recommendation of an inpatient DNR order. To our knowledge, no such study has been conducted in the cancer palliative care setting.
MATERIALS AND METHODS
We retrospectively reviewed medical records of consecutive patients with advanced cancer admitted to the inpatient palliative care service at The University of Texas M. D. Anderson Cancer Center between November 2002 and October 2008. The study protocol was approved by the institutional review board of The University of Texas M. D. Anderson Cancer Center.
Patients were considered for the study if it was determined by the study investigator while reviewing the patients' medical records that an inpatient DNR order discussion was appropriate. DNR discussion was considered to be appropriate for patients with metastatic or locally advanced refractory disease, or for patients with hematological malignancies who had not responded to antineoplastic therapy.
Patients were considered to have accepted DNR if a scanned inpatient DNR order was found in the chart or a dictated note indicated that a DNR order was in place. If patients had no written orders in the scanned documents or a dictated note indicating that a DNR order was in place, they were considered to have refused, because it is standard practice for the medical staff in our palliative care service to discuss DNR before transfer to the acute palliative care unit in all cases in which such discussion is deemed appropriate. Therefore, patients admitted to the acute palliative care unit without documentation of a DNR order were considered to have refused DNR status in this study.
Additional data were collected by chart review for 100 consecutive patients who refused DNR and for the first 200 patients who accepted DNR. The information collected included: demographic and clinical data, appropriateness for DNR discussion, referral statistics, timing and survival statistics, and advance directive status. Demographic data were as follows: age, sex, ethnicity, marital status, religious affiliation, insurance status, and profession. Clinical data included primary cancer diagnosis and date, disease stage at the time of admission, date of diagnosis of advanced cancer, and symptom burden at the time of acute palliative care unit admission as reported in the Edmonton Symptom Assessment Scale (ESAS).16 The ESAS is a widely used and validated tool to assess 9 symptoms (pain, fatigue, nausea, depression, anxiety, drowsiness, appetite, shortness of breath, sleep) and general feeling of well-being in a scale ranging from 0 (no symptoms) to 10 (worst symptom imaginable). Patients' ratings are recorded and graphed, and a total symptom distress score (0-90) is calculated as the sum of the first 9 symptoms' scores.16
Patient outcomes including inpatient death or place of discharge were also collected. Other information collected were date of inpatient DNR order, date of admission to the acute palliative care unit during the index hospitalization, source of palliative care referral (outpatient clinic, inpatient referral from other primary teams, or emergency center), type and date of first contact with our palliative care team, and dates for palliative care consultation during the current admission, hospital admission, discharge or inpatient death, and outpatient death (for outpatients, date of death was obtained from the Social Security Death Index). The number of days between diagnosis and first contact with the palliative care team, admission to the acute palliative care unit, and death was calculated, as was the number of days between diagnosis of advanced cancer and first contact with the palliative care team, admission to the acute palliative care unit, and death. The length of hospital and acute palliative care unit stay as well as days from discharge to death for those discharged alive were calculated. We also calculated the number of days from DNR discussion until death. For those patients who refused DNR and for whom a distinct discussion date was not available from the chart, the date of admission to the acute palliative care unit was considered to be the date of DNR discussion.
Regarding advance directives, the presence of each advanced directive (medical power of attorney, living will, and out of hospital DNR order) and the date when they were signed were documented, as was the date that each inpatient DNR order was implemented. The medical team responsible for implementing the inpatient DNR order was also identified.
Descriptive statistics were used to summarize the data. Chi-square tests were used to determine associations between categorical variables. Differences between continuous variables were analyzed using t tests for normally distributed data and Wilcoxon rank-sum tests for non-normally distributed data. Multivariate logistic regression analysis was performed with the variables found to be significantly associated with DNR refusal in univariate analysis. Recursive partitioning, which is a decision-tree method to identify specific groups of members of the population with greater probability of an outcome, was performed on a random selection of 200 patients to identify optimal patient classifications among significant factors identified in univariate models of DNR refusal. Once the optimal classifications were determined, the remaining 100 patients were used to estimate the odds of a DNR refusal between those classifications. This was done with the intent of generating preliminary data toward a predictive model that would allow the determination of patients at risk for complicated DNR discussions. Significance levels <.05 were considered statistically significant. Analyses were done using SPSS release 16.0 (SPSS, Chicago, Ill) and R version 2.3.1 (R Foundation for Statistical Computing, Vienna, Austria).
Of the 2538 medical records that were reviewed, 2530 (99%) had advanced cancer and were deemed appropriate for an inpatient DNR order discussion by the study investigators. The remaining 8 patients were considered appropriate to undergo CPR in the event of a cardiac arrest. Among the 2530 records reviewed, there were 2374 (94%) unique patients. Of these, 2274 (96%) accepted DNR, and 100 refused DNR (4%). Our final study population included 200 consecutive patients with a DNR order and 100 consecutive patients who refused it (Fig. 1). The majority of the DNR orders were written by the primary team (112 of 200, 56%) and the remaining by the palliative care team.
The demographic and clinical characteristics of the study population are summarized in Table 1, and the symptom burden of patients who accepted and refused DNR is shown in Table 2. DNR refusal was associated with greater ESAS pain and nausea scores, whereas acceptance was associated with higher ESAS shortness of breath scores.
Table 1. Demographic and Clinical Characteristics of Patients Who Accepted and Refused DNR
Significant variables in the univariate analysis were used for the multivariate regression analysis and recursive partitioning. Results of the multivariate logistic regression analysis are summarized in Table 3. African American race was a significant predictor of DNR refusal in the univariate but not in the multivariate model. The recursive partitioning model is illustrated in Figure 2 and Table 4. There was a lower rate of DNR refusal in patients who were seen by palliative care after the second day of hospital admission compared with those admitted directly to the acute palliative care unit for symptom control immediately after consultation with palliative care. Similarly, patients with longer admissions under the primary team (longer hospital admission to acute palliative care unit time) had higher rates of DNR acceptance.
According to the usual classification, pain was considered moderate when Edmonton Symptom Assessment Scale (ESAS) scores were between 4 and 7 and severe when ESAS scores were between 8 and 10. Dyspnea scores were categorized using the same strategy, and therefore it was considered mild when ESAS dyspnea scores were between 1 and 3.
Table 4. Recursive Partitioning Analysis: Validation Set
OR for Refusal of DNR
OR indicates odds ratio; DNR, do-not-resuscitate; CI, confidence interval.
Validation set logistic regression (96 remaining subjects, 34 refusals; patients in classifications III and VI were excluded from the model because they were not fit for predicting the outcome; 100% acceptance in group III and 100% refusal in group VI). R2 = 0.21. See Figure 2.
Table 5 summarizes the outcome of all 300 admissions. Among the 100 patients who refused DNR, 3 patients eventually underwent CPR. One died immediately after CPR, and 2 were discharged. The survival after CPR was 49 days in 1 case and not evaluable in the other (patient returned to home country after 22 days of admission). There were more inpatient deaths in those who accepted DNR compared with those who refused DNR.
Table 5. Patient Outcomes
Accept DNR, n=200
Refuse DNR, n=100
DNR indicates do-not-resuscitate.
Nursing home, skilled nursing facility or another hospital.
Table 6 summarizes the time intervals. Date of death for patients who were discharged alive was available for 177 (83%) of 212 patients. Median time between inpatient DNR and death was 15 days (interquartile range, 8-28). Median time from discharge to death was significantly shorter for those who had accepted DNR than for those who refused it (18 vs 85 days, P < .0001, Table 6). Time between DNR discussion and death was also shorter for patients who accepted DNR compared with patients who refused it. The median time between the first contact with the palliative care team and death was shorter in patients who accepted a DNR order than for those who refused it. For patients with DNR orders, median total hospital admission duration was significantly longer.
Table 6. Time Interval Between Events
Accept DNR, n=200, Median Days (IQR)
Refuse DNR, n=100, Median Days (IQR)
DNR indicates do-not-resuscitate; IQR, interquartile range; PC, palliative care; APCU, acute palliative care unit; PCC, palliative care consultation.
Patients discharged alive with subsequent death information, n=177.
Diagnosis to first contact with the PC team
Diagnosis to APCU admission
Advanced cancer diagnosis to first PCC
Advanced cancer diagnosis to APCU admission
Time from hospital admission to palliative consultation
Advance directives were present in 109 (55%) of 200 patients who accepted DNR and in 29 (29%) of 100 patients who refused it. Medical power of attorneys and living wills were found to be associated with acceptance of DNR (P = .004 and P = .002, respectively). Out-of-Hospital DNR orders were present in a small number of patients, 65 (22%) of 300, and were more common among patients who accepted DNR (54 of 65 vs 11 of 65; P = .002). The vast majority of living wills (68 of 75, 91%) exclusively provided directives to withdraw invasive treatments if the patient became unable to make decisions for him/herself, in the event of terminal/incurable disease. Six (8%) of 75 living wills provided directives to keep invasive treatments in all situations, and 1 (1%) of 75 requested that invasive treatments were to be provided in the event of terminal disease and withheld in the event of incurable disease. Presence of a medical power of attorney was significantly higher among Caucasians (71 [73%] of 97 vs 108 [53%] of 203 non-Caucasians, P = .001), Protestants (67 [70%] of 97 vs 110 [54%] of 203 for other religions, P = .05), and patients with administrative and management professions (35 [36%] of 97 vs 46 [22%] of 203, P = .02).
In this study, we found that the refusal rate for appropriately offered DNR orders in advanced cancer patients admitted to the acute palliative care unit was low, and that certain patient demographic and clinical characteristics were associated with DNR refusal. To our knowledge, there are no reports on the rate of DNR refusal among palliative care cancer patients under the care of specialized palliative care units. Physicians with experience in palliative care are less likely to recommend the initiation of CPR in patients similar to those represented in our study17 for various reasons. The diagnosis of advanced cancer is a poor prognostic factor for successful outcome of CPR,18-23 and it has also been associated with physical24, 25 and nonphysical harm.9 However, our group did not expect to see such a small DNR refusal rate, because we have noticed a high level of distress among members of the interdisciplinary palliative care team related to DNR refusal. This may have led to the perception that this occurred more frequently than in reality. More research is needed to understand the relation between DNR refusal and the degree of distress among the staff and to find better ways to support palliative care teams involved in the care of patients and families who refuse DNR.
Certain patient characteristics have been described in previous studies as associated with foregoing or choosing attempts for a CPR.26-28 Our finding suggesting that African American race is associated with DNR refusal is consistent with previous reports by other authors. This association is thought to result from cultural and religious beliefs regarding life prolonging measures, mistrust of the healthcare system in general, and fear of inadequate medical treatment that results from previous social inequities.28-36 However, in the multivariate model, African American race was not shown to be a significant predictor of DNR refusal. It may be that in the presence of other factors, particularly in the context of an interdisciplinary palliative care unit, the magnitude of the influence of ethnicity in predicting a DNR refusal becomes less important.
In a retrospective review of DNR utilization in 13,883 patients, Hanson and Rodgman reported that patients who were female, were Caucasian, had higher education and socioeconomic status, had poor functional status, or had reports of near death experience were most likely to have DNR orders.29 We report a similar finding for Caucasian patients. The finding that hematologic malignancy was associated with higher acceptance of DNR is of interest. We hypothesize that this is related to more advanced disease among patients with hematologic malignancy referred to the acute palliative care unit as compared with those with solid tumors. Previous studies have suggested that late referrals in these patients may be secondary to their lower symptom burden, more viable treatment options despite their advanced disease, and the absence of reliable indicators of advanced disease.37-39 Other authors have shown that the presence of a DNR order reflects the severity of the illness.40, 41
Patients' decision with regard to resuscitation was not associated with the total symptom burden at the time of admission to the acute palliative care unit. However, certain symptoms were found to be significantly associated with DNR acceptance and refusal. In our study, patients with moderate to severe dyspnea were more likely to have a DNR order in place. The role of dyspnea as a predictor of DNR refusal is not fully understood, but may have several plausible explanations. First, multiple discussions regarding prognosis and available treatment options may have occurred in this group of patients in anticipation of further decline in the clinical condition and use of more invasive treatment options such as intubation. This may have helped the patient and the family in deciding in favor of DNR. Second, there may be a perception of poor quality of life in patients with dyspnea. In a prospective observational study of 70 advanced cancer patients seen by a palliative care team, dyspnea was reported to interfere with mood, performance of normal daily activities, and enjoyment of life.42 Third, the presence of severe dyspnea may be perceived by patients and families as a sign of imminent death, causing resuscitative efforts to be more likely to be refused. Further research is needed to test these hypotheses.
There was a high DNR refusal rate in patients with severe pain in the multivariate analysis. It may be that the presence of intractable symptoms (other than dyspnea) is a hindrance to a more satisfactory discussion of end-of-life issues, and attempts to control them take precedence over end-of-life discussions. It is also possible that the expression of severe physical symptoms is related to patient's significant emotional distress,43, 44 causing less than satisfactory participation in DNR and other end-of life discussions. Further research is needed to test these hypotheses. If these findings are confirmed, repeated discussions or even postponing end-of life issues including resuscitation until symptom control is achieved should be part of the goals of care.
Patients who were admitted under their primary service longer, before their transfer to the acute palliative care unit, had lower rates of DNR refusals. We suspect that this group of patients had more opportunities to discuss their prognosis and treatment options before transfer to the acute palliative care unit compared with those directly admitted to the acute palliative care unit, who may not have had the opportunity for such a discussion with the primary service. Our findings suggest that DNR discussions do not frequently happen before the need for hospitalization occurs. This hypothesis would need to be confirmed in large clinical studies. If confirmed, this suggests that the primary team should be asked to participate in discussions regarding prognosis in all cases of direct acute palliative care unit admission.
Patients with at least 1 advance directive were more likely to have a DNR order in place, suggesting that end-of life and DNR discussions have possibly occurred and that patients have an understanding of goals of care enabling them to decide in favor of DNR.
With regard to homemaker status having a higher risk for DNR refusal, our initial hypothesis that this was related to being female and perhaps from an ethnic minority background was not confirmed after controlling for these factors in the multivariate analysis. Future research is recommended to test if this finding is consistent in this group of patients.
The R2 for the multivariate model of our study was of 0.23, suggesting that only 23% of the variation in DNR refusal can be explained by such model. Further research is needed to more completely characterize predictors of DNR refusals that could help in DNR and end-of-life discussions.
Patients with a DNR order were associated with in-hospital death and short survival compared with those who refused DNR. There are several possible explanations for this finding. The resulting inpatient death may simply reflect the advanced nature of the disease or the intervention provided to patients with a DNR order. Although the presence of a DNR order does not preclude patients from receiving other life-sustaining treatments, several studies have shown that the presence of a DNR order may result in lesser life-sustaining interventions18, 45, 46 and intensive care unit transfers.47 The reduced intensity of medical interventions during their acute palliative care unit stay and potentially after discharge may have contributed to our findings. However, we suspect that the advanced nature of the disease is the primary determinant of the number of inpatient deaths in those who had a DNR order. Further research needs to be done to test this hypothesis, which if confirmed would have implications in the how the medical staff provide informed consent for DNR.
The timing of a DNR order and other aspects of advance care planning is crucial for the patient and family.48, 49 Most discussions and DNR orders are placed within 0 to 3 days before death, when prognosis is poor, treatment options are limited, and death is imminent.27, 35, 50 In a study of Korean patients, Oh et al reported that 87% of patients had a DNR order signed 8 days before death, and about 10% were signed on the day of death.51 In our study, the DNR order was placed around 15 days before death, which is earlier than what has been previously reported in the literature.27, 35, 50, 51 We hypothesize that the longer interval might be directly related to the palliative care team intervention itself or to an increasing awareness about the importance of resuscitation status by all cancer specialties, leading to earlier DNR discussions.
The survival after the diagnosis of advanced cancer in this population is short, and has previously been described by our group as in the range of 8 months,52 a finding replicated in the current study (11-13 months). This relatively short estimated survival justifies the importance of having a meaningful advance care plan in place that could be documented in the form of advance directives. Our findings are similar to those in the literature that showed higher rates of advance directive completion in Caucasians.53 It is believed that the explanation for this difference in advanced directives is similar to those cited for DNR utilization, such as mistrust of the healthcare system, health disparities, cultural perspective on death and suffering, and unique family dynamics that may be culturally based.54-56
There have been recent discussions around the wording of the term “do not resuscitate.” Several authors have suggested that this term might elicit negative reactions or be too tightly related to withdrawal of care.5-7 However, the suggested new term “allow natural death” still has its opponents. For example, it is hypothesized that a change in name will not be able to drive a change in practice, and there are concerns about the finding that the new suggested term is actually even more vague.57, 58 There are also concerns related to the finding that painful conversations about end-of-life care should not be avoided, because they can end up causing more suffering and harm to patients and families.58 In our study, we kept the term “do not resuscitate” to reflect the still current nomenclature and also the standard wording used in our service. We expect that further research will help better determine whether either term is able to bring greater success in the care of these patients and their families.
As inherent to the retrospective methodology, this study has limitations. To obtain our sample of 100 DNR refusal patients, data collection for this group needed to span through charts of patients seen across several years (2002-2008), whereas data for the 200 patients who accepted DNR were collected for patients at the end of the period only, because of limited resources. One possible concern would be a change in attitudes and beliefs by patients in the last 6 years. We are not aware of any data showing that such a change in attitudes has taken place. However, our findings at this point should be considered preliminary, and more research is needed to confirm them. In addition, future studies could select and match patients with certain ethnic, demographic, and socioeconomic characteristics for the purpose of comparison in an effort to better determine the main influences on the outcomes of DNR discussions. Unfortunately, we were not able to obtain data regarding the refusal of DNR orders among patients not transferred to the palliative care service.
Inpatient DNR orders were successfully obtained in the vast majority of hospitalized patients for whom DNR discussion was considered to be appropriate. There is a significant proportion of patients who refused DNR despite discussions with the palliative care team. Patients who refused DNR seem to be a source of distress to the palliative care team and can be a source of financial burden to families, hospitals, and third-party payers. This study identifies possible predictors of complicated DNR discussions that include time from admission to palliative care consultation, time from hospital admission to transfer to the acute palliative care unit, symptoms such as pain and nausea, and absence of advance directives. Advance directives were present in nearly half of the patients who accepted DNR and in <30% of those who refused DNR.
We thank Ray Chacko and Valerie Poulter for invaluable help with data collection.
CONFLICT OF INTEREST DISCLOSURES
Eduardo Bruera is supported in part by National Cancer Institute R01 grants CA122292-01 and CA124481-01, and National Institute of Nursing Research grant NR010162-01A1. David Hui is supported by a grant from the Clinician Investigator Program, Royal College of Physicians and Surgeons of Canada.