Goals of Care and End-of-Life Decision Making for Hospitalized Patients at a Canadian Tertiary Care Cancer Center

Article Outline

• Abstract
• Introduction
• Methods
  • Patients and Setting
  • Frequency and Appropriateness of Investigations
  • Cancer Treatments
  • Goals of Care Milestones
  • Statistical Analysis
• Results
  • Baseline Characteristics on Admission
  • Frequency and Appropriateness of Investigations
  • Cancer Treatments
  • Goals of Care Milestones
  • Correlation Between Goals of Care Milestones and Clinical Management
  • Practice Pattern Differences by Discipline
• Discussion
• Acknowledgments
• References
• Copyright

Abstract 

Limited information is available regarding the quality of end-of-life care at cancer centers. We sought to characterize the end-of-life decision-making process for advanced cancer patients admitted to our tertiary cancer center, and to examine the association between goals of care and practice patterns. Information on patient characteristics, investigations, cancer treatments, and goals of care was collected retrospectively for consecutive patients who died at the inpatient unit of the Vancouver Cancer Center between January 1, 2005 and December 31, 2006. One hundred eighteen advanced cancer patients had a median admission duration of 10 days (range 1–64 days). A median of two tests per day was performed, with a decreasing trend over time (P<0.001). Forty percent received cancer treatments during hospitalization, with 75% terminated prematurely. Do-not-resuscitate orders, supportive care plans, and diagnosis of dying were documented for 96%, 86%, and 76% of the patients, respectively. Early establishment of supportive care plan and diagnosis of dying were associated with timely discontinuation of cancer treatments (Spearman coefficients 0.47 and 0.60, respectively). Multivariate analysis revealed that timely diagnosis of dying was associated with early establishment of code status (P=0.042), supportive care plans (P<0.001), and discontinuation of cancer therapy (P=0.005). Cancer patients who died at our oncology center were investigated and treated intensively during their short hospitalization. Early establishment of goals of care may be associated with changes in practice consistent with improved quality of care.

Key Words: Advanced cancer, goals of care, investigations, cancer treatments, quality of care, end of life

Back to Article Outline

Introduction 

Patients with advanced cancer are bombarded with numerous decisions involving treatments, goals of care, and end-of-life planning from the time of diagnosis until death. Many of these decisions are highly complex, emotionally charged, and have significant impact on how patients are managed. One of the key roles of the health care team is to help guide patients through the maze of difficult choices by providing individualized recommendations, taking into account the patient's preferences, disease state, treatment options, and resources.

The role of clinical decision making is particularly important for advanced cancer patients dying in acute care settings. These patients typically suffer from significant physical, psychological, and existential distress, and are sometimes too ill to engage in discussions regarding health care decisions. This is further complicated by the acute care culture of intense investigations and treatments to diagnosis and treat any potentially reversible causes.¹ These actions often result in significant delays in decision making regarding end-of-life care, leading to aggressive interventions such as intubation and chemotherapy,¹ with appropriate comfort measures being neglected.

A number of studies have specifically examined the hospital course and end-of-life decision-making process for hospitalized patients dying in acute care settings. The landmark SUPPORT Phase I study examined 4,301 hospitalized patients with advanced diseases, and revealed multiple distressing findings.² Less than half of the physicians knew about their patients' preferences for avoidance of cardiopulmonary resuscitation (CPR), and 46% of do-not-resuscitate (DNR) orders were written within only two days of death. Other studies focusing on patients who died in acute care hospitals also documented aggressive measures and significant delays in establishing goals of care,³, ⁴, ⁵, ⁶, ⁷ in contrast to patients in nursing homes and palliative care units.⁷, ⁸

Although it makes intuitive sense that improved communication and early establishment of goals of care can lead to improved quality of care, past studies on this issue have resulted in conflicting findings. Phase II of the SUPPORT study involved 4,804 hospitalized patients randomized to either usual medical care or longitudinal interventions aimed to improve communication and decision making by providing timely and reliable prognostic information. This study showed no difference in patient-physician communication or in the five targeted outcomes, including incidence or timing of written DNR orders, physicians' knowledge of their patients' preferences not to be resuscitated, number of days spent in an intensive care unit (ICU), comatose or receiving mechanical ventilation before death, level of reported pain, or hospital resource use.² However, a recent study of advanced cancer patients demonstrated that end-of-life discussions were associated with lower rates of ventilation, resuscitation, ICU admissions, and earlier hospice enrollment.⁹ Of note, these studies were carried out over weeks to months prior to patients' death, and much is still not known about the relationship between establishment of goals of care and the actual delivery of care during the last admission. A more complete understanding of this process can help us minimize aggressive interventions and optimize quality of care.

We hypothesized that advanced cancer patients admitted to our tertiary care cancer center were investigated and treated aggressively at the end of life, and that early establishment of goals of care during the hospital stay could help to minimize inappropriate interventions. The objective of this study was to characterize the quality of care provided to cancer patients who died at the Vancouver Cancer Center, specifically related to establishment of goals of care milestones, investigations, and cancer treatments. We also sought to examine the association between goals of care milestones and management practices.

Back to Article Outline

Methods 

Patients and Setting 

All the patients who died on the inpatient unit of the Vancouver Cancer Center between January 1, 2005 and December 31, 2006 were included in this study. The Vancouver Cancer Center is a tertiary care oncology center serving a catchment population of 4.2 million in the province of British Columbia, Canada. It has an inpatient unit of 30 beds, with no ICU or palliative care beds. The patients who require critical care are sent to an adjacent general hospital, whereas the patients who cannot return home following admission and who require palliative care are transferred to community palliative care units or hospices as appropriate, when possible.

Information regarding patient demographics, cancer and admission diagnosis, clinical status on admission, frequency and appropriateness of investigations, active cancer treatments, disease trajectory, causes of death, and goals of care was retrospectively reviewed by a medical oncologist (D. H.). A random sample of 10 charts was further reviewed by a palliative care specialist (P. H.) to ensure accuracy of data collection. Supportive care plan, diagnosis of dying, and appropriateness of investigation data were available for all but three patients because of missing charts. The Research Ethics Board at the British Columbia Cancer Agency approved this study and waived the requirement for informed consent.

Frequency and Appropriateness of Investigations 

Investigations performed during the final admission were documented under the categories of blood tests, microbiology, imaging, and others (e.g., electrocardiogram and arterial blood gas). Procedures requiring radiologic interventions (e.g., ultrasound-guided paracentesis) were classified under imaging tests. Investigations were numerated based on how they are typically ordered by physicians. For instance, complete blood count (CBC), electrolytes/urea/creatinine, calcium/magnesium/phosphate, liver function test (LFT), troponin/creatine kinase, urinalysis, blood culture, urine culture, chest X-ray, computed tomography chest, magnetic resonance head, bone scan, electrocardiogram (ECG), and arterial blood gas (ABG) were each considered an individual test.

In addition to the frequency of investigations, we also reviewed the progress notes in detail to evaluate the proportion of tests for which there was a clear indication (diagnosis, treatment, or monitoring), balancing the relative invasiveness to determine whether the tests would be repeated again if the reviewers were put in the same clinical situation. Based on this definition, we assigned investigations during hospitalization into one of five categories of appropriateness: 1=0%–20%, 2=21%–40%, 3=41%–60%, 4=61%–80%, and 5=81%–100%. For instance, a patient with 10 CBCs performed during a two-week hospital stay may have 61%–80% of tests judged to be appropriately contributing to the patient's care for a clear indication (e.g., upper gastrointestinal (GI) bleed and a need for transfusions), whereas another patient with the same number of tests but the majority with no clear justification may be assigned a score of 21%–40%. A limitation of this measure is its subjectivity in interpretation of what constitutes appropriateness. Thus, we characterized the reliability of this measure in which a second independent observer (P. H.) determined the appropriateness of investigations (i.e., CBCs, electrolytes, LFTs, imaging, and overall) for 10 patients selected at random. The interrater reliability was moderate, with a Cohen's kappa of 0.70. The concordance rate (±1 category) was 80%, with identical values assigned to 54% of all cases. The discordant rate (≥2 category difference) was 10%. No patient had a ≥3 category difference.

Cancer Treatments 

We documented the start and end dates for the last cancer treatment administered. Systemic therapies were classified as chemotherapeutic agents, hormones, and targeted therapies. None of the patients were on any experimental therapies or clinical trials. External beam radiation was the only type of radiotherapy administered. In this study, early termination of treatment was defined as less than two cycles of chemotherapy/targeted agent since hospitalization, or unable to complete the planned course of radiation.

Goals of Care Milestones 

Goals of care were defined in this study as (1) the presence of a DNR order in the chart, (2) whether a comfort care plan was adopted, and (3) whether a clinician considered a patient to be dying. They were assessed based on three validated questions previously developed for this purpose.⁴ A comfort care plan was noted to be in place if the following phrases were present: “comfort care,” “palliative care,” “supportive care,” “hospice,” or “palliative care unit.” Evidence that a patient was considered dying was based on documentation of the following language in the progress notes: “end stage,” “dying,” “terminally ill,” “moribund,” “hopeless,” “grim/grave prognosis,” or “days/weeks of life.” The dates on which the milestones were first documented were recorded.

Statistical Analysis 

The Statistical Package for the Social Sciences (SPSS) software was used for statistical analysis (version 14.0, SPSS Inc., Chicago, IL). Comparisons were made using the Student's t-test and ANOVA for independent samples for continuous outcomes, and Pearson's Chi-square test for categorical outcomes. We applied both linear and Poisson regression modeling to determine the trends regarding frequency of investigations over the admission. The Spearman's correlation test was used to examine the relationship between nonparametric variables.

To identify factors that were independently associated with the timing of establishment of goals of care milestones, appropriateness of clinical investigations, and the timing of discontinuation of cancer treatments, we performed nonparametric multivariate linear regression using backward elimination. Each model incorporated age, gender, attending service, cancer stage, Eastern Cooperative Oncology Group (ECOG) performance status, frequency of investigation, appropriateness of investigation, length of hospital stay, and the duration between admission and establishment of milestones (excluding the dependent variable). A two-tailed P-value of less than 0.05 was considered to be statistically significant.

Back to Article Outline

Results 

Baseline Characteristics on Admission 

During the two-year study period, 118 (4.7%) patients died out of a total of 2,533 admissions. The characteristics of the 118 patients who died in hospital are shown in Table 1. Nine percent were diagnosed with hematologic malignancies and 91% had solid tumors, of whom 84% had metastatic disease. Notably, none of the patients who died in hospital had early stage disease. Gastrointestinal (31%), lung (18%), and breast (12%) malignancies accounted for the top 60% of all admissions in this cohort. All the patients were symptomatic on admission and required supportive care, with 41% admitted for palliative anticancer therapy and 6% admitted for workup of malignancy, as a cancer diagnosis had not been established at the time of admission. Pain (61%), anorexia and cachexia (31%), dyspnea (30%), and nausea and vomiting (29%) were common symptoms on admission. Ninety-one percent had an ECOG performance status of three or greater.

Table 1. Characteristics of Cancer Patients Who Died in Hospital, by Medical Specialty

Clinical Characteristics	Medical Oncologist (n=80)	Radiation Oncologist (n=28)	Gynecologist (n=10)	P-valuea	Total (n=118)
Median age	61	70	57	0.012	62 (19–89)
Female gender	45	14	10	NS	69 (58.5%)
Symptoms
Dyspnea	23	7	4	NS	34 (29.8%)
Confusion	14	7	0	NS	21 (18.4%)
Pain	42	22	6	0.037	70 (61.4%)
Nausea and vomiting	22	6	5	NS	33 (28.9%)
Anorexia/cachexia	26	7	2	NS	35 (30.7%)
Bleeding	7	3	0	NS	11 (9%)
ECOG 3–4	69	25	9	NS	108 (91%)
Reason for admission
Symptom management	80	28	10	NS	118 (100%)
Palliative anticancer tx	27	17	4	0.016	48 (40.7%)
Diagnostic workup	5	1	1	NS	7 (5.9%)
Cancer diagnosis
Gastrointestinal	29	7	0	—	36 (30.5%)
Lung	10	11	0	—	21 (17.8%)
Breast	13	1	0	—	14 (11.9%)
Hematologic	11	0	0	—	11 (9.3%)
Gynecological	1	1	8	—	10 (8.5%)
Genitourinary	6	3	0	—	9 (7.6%)
Primary unknown	6	0	2	—	8 (6.8%)
Head and neck	3	4	0	—	7 (5.9%)
Others	1	1	0	—	2 (1.7%)
Stage
Solid tumor—metastatic	63	18	9	—	90 (76.3%)
Solid tumor—locally advanced	6	10	1	—	17 (14.4%)
Hematologic—advanced	11	0	0	—	11 (9.3%)

ECOG=Eastern Cooperative Oncology Group; NS=not significant; tx=treatment.

The median duration from diagnosis to admission was 10 months (range −0.1 to 204.2 months). However, 14% had the diagnosis of cancer made less than one month prior to the final admission and 6% during the last hospitalization, suggesting a highly aggressive course of disease for these individuals. The median duration of admission was 10 days (range 1–64 days), with 14% of the patients admitted for three days or less before their deaths. Ten percent of the patients had a prolonged course in hospital lasting more than four weeks.

Frequency and Appropriateness of Investigations 

In total, 3,218 investigations were performed during a combined 1,568 hospital days, with an average of two tests per day for every admitted day (Table 2). Blood work (75%) constituted most of the investigations, with CBC (25%) and electrolytes (25%) ordered most frequently.

Table 2. Frequency of Investigations in 115 Cancer Patients Who Died in Hospital^a

Investigations	Number of Tests (% of Total)	Number of Tests per Day	Median (Range) per Admission
Blood work	2418 (75.1%)	1.54	12.5 (0–153)
CBC	787 (24.5%)	0.50	4 (0–60)
Electrolytes	803 (25.0%)	0.51	4 (0–62)
Liver function tests	368 (11.4%)	0.23	1 (0–22)
INR	236 (7.3%)	0.15	1 (0–15)
Others	224 (7.0%)	0.14	1 (0–18)
Microbiology	339 (10.5%)	0.22	1 (0–20)
Imaging	307 (9.5%)	0.20	2 (0–10)
CXR	142 (4.4%)	0.09	1 (0–7)
Others	165 (5.1%)	0.11	1 (0–9)
Others	154 (4.8%)	0.10	1 (0–11)
Total	3218 (100%)	2.05	17 (0–179)

CBC=complete blood count; CXR=chest X-ray; INR=international normalized ratio.

The number of tests per day inversely correlated with the duration of admission (Spearman's correlation coefficient −0.31, P=0.001). Using two different regression models, there was a decreasing trend of the frequency of investigations (Poisson regression slope −0.137, standard error (SE)=0.017, P<0.001, Fig. 1) over the hospital course. However, investigations were done in 91% and 28% of the patients during the last week and last day of life, respectively.

• 
  • View Large Image
  • Download to PowerPoint

• Fig. 1 

  Regression models of investigation frequency during hospitalization. The number of investigations per week was plotted for 118 patients over the hospital course. Both linear regression and Poisson regression models revealed a significant trend of decreasing frequency of investigations over time.

We also evaluated the appropriateness of investigations, based on the proportion of tests for which there was a clear indication (Fig. 2). In 41% of the patients, >80% of investigations were considered appropriate, whereas in 75% of the patients, >60% of the tests were justified. LFTs and imaging investigations were more frequently performed appropriately, whereas CBC and electrolytes tended to be ordered with less discretion (P<0.001, Chi-square test).

• 
  • View Large Image
  • Download to PowerPoint

• Fig. 2 

  Appropriateness of investigations based on the type of investigation. The appropriateness of investigations was a measure of the proportion of tests for which there was a clear indication (diagnosis, treatment, or monitoring). Imaging investigations and liver function tests were associated with a higher percentage of patients with higher appropriateness scores compared to CBC and electrolytes.

Cancer Treatments 

Among the 118 patients who died in hospital, 17 (14.4%) never received any cancer treatments. Forty-two (36%) patients received active treatments within the last two weeks of their lives, including 21 (18%) treated with chemotherapy (Table 3).

Table 3. Cancer Treatments Given to 118 Cancer Patients Who Died in Hospital

Anticancer Therapy	Number of Patients (%)
Last active cancer treatment before or during admission
Chemotherapy	61 (51.7%)
Radiation therapy	33 (28.0%)
Hormonal therapy	3 (2.5%)
Biological therapy (Phase I trial)	1 (0.8%)
Chemoradiation	1 (0.8%)
Chemoembolization	1 (0.8%)
Surgery (laparotomy)	1 (0.8%)
None	17 (14.4%)
Active cancer treatment during admission
Palliative radiation	23 (19.5%)
Palliative chemotherapy	22 (18.6%)
Hormonal therapy	2 (1.7%)
Palliative chemoradiation	1 (0.8%)
None	70 (59.3%)
Active cancer treatment within 14 days of deatha
Palliative radiation	19 (14.4%)
Palliative chemotherapy	21 (17.8%)
Hormonal therapy	1 (0.8%)
Palliative chemoradiation	1 (0.8%)
None	76 (64.4%)
Early termination of treatments during admission
Radiation	13 of 23 (56.5%)
Chemotherapy	22 of 22 (100%)
Chemoradiation	1 of 1 (100%)
Hormonal therapy	0 of 2 (0%)

During the last admission, 25 (21%) and 24 (20%) patients were treated with systemic therapy and radiation, respectively. All chemotherapy treatments given during admission were terminated prematurely, whereas 48% of the patients who received radiation during the last admission were able to complete the planned course. Among patients who received cancer therapy in hospital, the median duration between last chemotherapy and death was 7.5 days, and the median time between last radiation and death was five days (Fig. 3). Among patients who completed the course of palliative radiation during admission, the median duration between last radiotherapy fraction and death was 12.5 days (interquartile range [Q1–Q3] 6–18 days), with only one patient living four weeks post-treatment. Given the short interval between radiation and death, few patients in our cohort likely benefited from this treatment.

• 
  • View Large Image
  • Download to PowerPoint

• Fig. 3 

  Decision milestones from date of death. Box plot displaying the number of days from death to admission, last chemotherapy treatment (only for patients who received chemotherapy during admission), last dose of radiation (only for patients who received radiation during admission), last investigation, and establishment of goals of care including DNR, supportive care plan, and diagnosis of dying. The numbers indicate the median number of days from each of the respective decision points to the date of death.

Goals of Care Milestones 

DNR orders, supportive care plans, and diagnosis of dying were documented for 94%, 86%, and 76% of the patients, respectively (Table 4). The median time from establishment of these key indicators to death was eight, seven, and four days (Fig. 3). The diagnosis of dying was consistently documented later and in fewer patients than the other two milestones (Table 4).

Table 4. Number of Patients with Goals of Care Milestones Established, by Discipline and Duration of Admission

Code status was addressed in 112 (94%) patients, with DNR orders written for 111 of them, and one patient requesting full code despite repeated discussions. One of the patients with a DNR order in place mistakenly received CPR. Among the six patients without documentation of code status discussion, five received some form of resuscitation, with CPR performed on three of them.

Correlation Between Goals of Care Milestones and Clinical Management 

The timing of establishment of the three goals of care milestones (DNR orders, supportive care plans, and diagnosis of dying) correlated with each other, with a particularly strong association between diagnosis of dying and establishment of supportive care plan (Table 5). Early establishment of goals of care correlated with more appropriate ordering of investigations. The timing of establishment of supportive care plans and diagnosis of dying also were associated with discontinuation of cancer treatments (Table 5). These milestones also correlated with length of stay. The patients with longer hospitalizations were more likely to have had goals of care established later in admission.

Table 5. Univariate Analysis for Timing of Establishment of Goals of Care Milestones^a

Timing of Goals of Care Milestones	Goals of Care Milestones
	Timing of Code Status (Days)	Timing of Care Plan (Days)	Timing of Dying Diagnosis (Days)
Timing of code status (days from admission)	—	0.39	0.49
		P<0.001	P<0.001
Timing of care plan (days from admission)	0.39	—	0.70
	P<0.001		P<0.001
Timing of dying diagnosis (days from admission)	0.49	0.70	—
	P<0.001	P<0.001
Length of admission (days)	0.30	0.49	0.65
	P=0.001	P<0.001	P<0.001
Appropriateness of investigations	−0.25	−0.20	−0.26
	P=0.008	P=0.048	P=0.014
Timing of stopping treatment (days from admission)	0.10	0.47	0.60
	P=0.52	P=0.002	P<0.001

—=not applicable.

In multivariate analysis, the timely establishment of both code status and supportive care plans were associated with early diagnosis of dying (Table 6). The diagnosis of dying, in turn, was associated with early discontinuation of cancer therapy. Furthermore, the appropriateness of clinical investigations improved with both early discontinuation of cancer treatments and decreased frequency of investigations (Table 6).

Table 6. Multivariate Analysis for Timing of Establishment of Goals of Care Milestones and Practice Patterns^a

Variables	Multivariate Analysis
	P-valuea
Early establishment of code status
Timing of diagnosis of dying	0.042
Early establishment of care plan
Timing of diagnosis of dying	<0.001
Length of admission	0.037
Early diagnosis of dying
Timing of care plan	<0.001
Timing of discontinuation of cancer treatment	0.045
Appropriateness of investigations
Frequency of investigations	<0.001
Timing of discontinuation of cancer treatment	0.015
Early discontinuation of cancer treatments
Timing of diagnosis of dying	0.005
Length of admission	0.028

Practice Pattern Differences by Discipline 

We compared the admission characteristics, investigations, treatments, and goals of care milestones among medical, radiation, and gynecologic oncologists. The patients admitted under radiation oncology tended to be older (P=0.01), required pain control (P=0.04), and were admitted for palliative cancer treatments (P=0.02, Table 1). Medical oncology patients had more investigations on average than those under radiation and gynecologic oncology (2.6±1.8 vs. 1.8±0.8 and 1.8±1.0 investigations per day, respectively); however, the appropriateness of investigations did not differ among the teams, suggesting that medical oncology patients may inherently require more testing because of the nature of their diseases. No significant differences were detected among the three disciplines regarding cancer treatments and establishment of goals of care milestones.

Back to Article Outline

Discussion 

We found that cancer patients who died at our oncology center were investigated and treated intensively during their short hospitalizations. Our results suggest that there is a lot of room for improvement in the delivery of quality end-of-life care on our unit. Quality of care for advanced cancer patients is dependent on a number of factors, including the establishment of goals of care, the intensity of investigations and cancer treatments, pain and symptom management, and psychosocial support, including advance care planning, communication, and availability of palliative care resources, such as hospices.¹⁰ Although a number of studies have examined the quality of care for dying patients in general hospitals,², ³, ⁴, ⁵, ⁶, ⁷, ¹¹ our study represents one of the few to specifically examine the unique challenges for advanced cancer patients admitted to a tertiary care oncology center. The balance between quality-of-life measures and life-prolonging treatments presents a logistic and philosophical dilemma for patients who are dying while admitted to a cancer center and for those caring for them, especially when the goals of care have not been clearly established.

This is the most comprehensive study to date examining both the frequency and appropriateness of investigations in cancer patients. We observed an overall decreasing trend in the number of investigations over the course of hospitalization; however, 28% of the patients still had tests performed on the very last day of life, despite the fact that 78% of the patients had the diagnosis of dying made a median of four days prior to death. Although some of these tests might be justified as physicians searched for reversible causes of clinical deterioration, it was clear that many did not contribute to the patient's care and were thus unnecessary.¹² Although there will inevitably be an occasional patient who dies unexpectedly, for example, of pulmonary embolus or myocardial infarction, improving the ability of clinicians to recognize patients who are actively dying and to differentiate them from patients who may recover would help to minimize unnecessary investigations and treatments.

We also introduced the concept of appropriateness of investigation in this study as an attempt to quantify this by the nature of subjective measure. The inter-rater reliability between two independent observers was moderate. This outcome measure has face validity. For instance, imaging investigations and LFTs were more often ordered appropriately than CBCs and electrolytes, which is consistent with clinical practice. In multivariate analysis, the appropriateness of investigation was associated with both the frequency of testing and early discontinuation of cancer therapy, which highlights the link between treatment and the need for frequent monitoring. Further studies are required to determine the validity and reliability of this measure.

A somewhat distressing finding is that among patients with far advanced cancer, 38% received aggressive treatments within the last two weeks of life and 41% during the last hospitalization. Given the fact that the great majority of these patients had poor performance status (ECOG≥3), treatment was probably contraindicated in many. Indeed, therapy had to be terminated prematurely for about half of all the patients on radiation, and for all the patients on chemotherapy. Although it is not unusual for patients to be admitted or transferred to cancer centers with the intention of starting antineoplastic treatments, oncologists carry the important responsibility of deciding whether therapy is warranted and/or contraindicated, after assessing the patients. For the particular cohort in this study, therapy was not only futile, but could be potentially harmful. Indeed, the literature is filled with studies documenting the increasing trends of chemotherapy use close to the end of life,¹³, ¹⁴ with hope being a key reason behind both patients'¹⁵, ¹⁶ and oncologists'¹⁷ decision to pursue aggressive therapies. Improved prognostication tools that help clinicians to identify patients who are “actively dying,” coupled with treatment decision tools, can help limit the inappropriate use of potentially harmful therapies.

Compared to previously published studies of hospitalized patients in general hospitals,⁴, ⁵, ⁶ goals of care milestones were established more frequently at our institution, although still suboptimally. Documentation of the diagnosis of dying is arguably the most important of the three, as it signals to the health care team that the patient is close to the end of life, yet it was established later and less frequently compared to DNR and supportive care plans in our cohort. In some cases, it may be clear in the clinician's mind that the patient is dying, but it has not been written in the chart; we argue that documentation is essential for establishment of this important diagnosis and for communication among team members. The delay in diagnosis of dying is probably related to the technical difficulty and emotional hesitancy in establishing a “terminal” diagnosis. Indeed, no clear definition for “dying” exists. Nevertheless, symptoms such as poor performance status, anorexia-cachexia, inability to swallow, dyspnea, and delirium are suggestive of this diagnosis in patients with advanced cancer. It is important to recognize that the diagnosis of dying can be made knowing that there is still a small chance of recovery in some patients.¹⁸, ¹⁹ Indeed, a frank discussion with patients and their families would provide an important opportunity to plan ahead, with the potential for improving the quality of end-of-life care.²⁰

To our knowledge, this is the first study to demonstrate that early establishment of goals of care milestones, particularly diagnosis of dying, was associated with changes in clinical practice consistent with improved quality of care. However, the strength of correlation was only mild to moderate overall, suggesting the multifactorial nature and complexity of clinical decision making. Importantly, the goals of care milestones remained significant after accounting for length of stay and other clinical factors with multivariate analysis. Our finding is supported by a recent study that demonstrated that the end-of-life discussions were associated with less aggressive medical care near death and earlier hospice referrals.⁹

We further analyzed the differences in practice patterns between medical oncologists, radiation oncologists, and gynecologic oncologists. The patients admitted under medical oncology had more investigations, although no difference in appropriateness of investigation was found. This finding may suggest that medical oncology patients required more testing because of the intrinsic nature of their diseases, such as acute renal failure, liver dysfunction, and infections. Importantly, it also distinguishes appropriateness of investigation as a measure that is distinct from frequency of investigation. We found no significant differences among the three disciplines in regard to treatments or goals of care milestones, although this would need to be interpreted with caution given the small sample size.

Limitations of this study include its retrospective nature and the small sample size from a single institution. First, the patients who were transferred to other hospitals or hospices and died shortly after were not included in this study; nevertheless, the proportion of patients who fit this profile is estimated to be low, as oncologists in our institution generally avoid transferring patients who are deemed to be imminently dying. For instance, there were only 18 transfers to inpatient hospices over the two-year study period. Second, we only collected data from the last admission, and did not include prior information on goals of care milestones or treatments. Like many other institutions, however, our center requires a revisit of goals of care on each admission, recognizing that the patient's clinical status and preferences evolve rapidly over time. Thus, the information from the last admission would be most representative for the purpose of this study. Third, other important aspects of quality of end-of-life care, such as symptom assessment and psychosocial interventions, were not included in this study. Fourth, the use of “appropriateness of investigation” as an indicator of quality of care warrants further validation. Finally, comparison with a group of advanced cancer patients who were discharged alive from our cancer center would be informative.

Findings from this study have important implications for the management of cancer patients at the end of life. On the basis of our data, we conclude that the cancer patients who died at our tertiary care cancer center were investigated and treated intensively during their short hospitalizations, suggesting that there remains much room for improvement in optimizing the quality of care. Importantly, this study demonstrated that early recognition of the dying process may help to limit aggressive measures. To help patients and families navigate through times of uncertainties, clinicians need to be skilled at prognostication and communication, providing realistic expectations to facilitate discussions surrounding goals of care and end-of-life issues. Further developments in the areas of prognostication models, treatment decision tools for clinicians, decision-making aids for patients, and integrated care pathways can help enhance the quality of care for advanced cancer patients.

Back to Article Outline

Acknowledgments 

We would like to thank Mr. Ryan Wood for his assistance in statistical analysis.

Back to Article Outline

References 

1. Harrington SE, Smith TJ. The role of chemotherapy at the end of life: “when is enough, enough?”. JAMA. 2008;299:2667–2678
   • View In Article
   • CrossRef
2. A controlled trial to improve care for seriously ill hospitalized patients. The study to understand prognoses and preferences for outcomes and risks of treatments (SUPPORT). the SUPPORT principal investigators. JAMA. 1995;274:1591–1598
   • View In Article
   • MEDLINE
3. Bauduer F, Capdupuy C, Renoux M. Characteristics of deaths in a department of oncohaematology within a general hospital. A study of 81 cases. Support Care Cancer. 2000;8:302–306
   • View In Article
   • MEDLINE
   • CrossRef
4. Fins JJ, Miller FG, Acres CA, et al. End-of-life decision-making in the hospital: current practice and future prospects. J Pain Symptom Manage. 1999;17:6–15
   • View In Article
   • Abstract
   • Full Text
   • Full-Text PDF (128 KB)
   • CrossRef
5. Middlewood S, Gardner G, Gardner A. Dying in hospital: medical failure or natural outcome?. J Pain Symptom Manage. 2001;22:1035–1041
   • View In Article
   • Abstract
   • Full Text
   • Full-Text PDF (174 KB)
   • CrossRef
6. Becker G, Sarhatlic R, Olschewski M, et al. End-of-life care in hospital: current practice and potentials for improvement. J Pain Symptom Manage. 2007;33:711–719
   • View In Article
   • Abstract
   • Full Text
   • Full-Text PDF (112 KB)
   • CrossRef
7. Sato K, Miyashita M, Morita T, et al. Quality of end-of-life treatment for cancer patients in general wards and the palliative care unit at a regional cancer center in Japan: a retrospective chart review. Support Care Cancer. 2008;16(2):113–122
   • View In Article
   • CrossRef
8. Tse DM, Chan KS, Lam WM, Leu K, Lam PT. The impact of palliative care on cancer deaths in Hong Kong: a retrospective study of 494 cancer deaths. Palliat Med. 2007;21:425–433
   • View In Article
   • CrossRef
9. Wright AA, Zhang B, Ray A, et al. Associations between end-of-life discussions, patient mental health, medical care near death, and caregiver bereavement adjustment. JAMA. 2008;300:1665–1673
   • View In Article
   • CrossRef
10. Earle CC, Park ER, Lai B, et al. Identifying potential indicators of the quality of end-of-life cancer care from administrative data. J Clin Oncol. 2003;21:1133–1138
    • View In Article
    • MEDLINE
    • CrossRef
11. Goncalves JF, Goyanes C. Use of chemotherapy at the end of life in a Portuguese oncology center. Support Care Cancer. 2008;16:321–327
    • View In Article
    • CrossRef
12. Asola R, Huhtala H, Holli K. Intensity of diagnostic and treatment activities during the end of life of patients with advanced breast cancer. Breast Cancer Res Treat. 2006;100:77–82
    • View In Article
    • MEDLINE
    • CrossRef
13. Earle CC, Neville BA, Landrum MB, et al. Trends in the aggressiveness of cancer care near the end of life. J Clin Oncol. 2004;22:315–321
    • View In Article
    • MEDLINE
    • CrossRef
14. Earle CC, Landrum MB, Souza JM, et al. Aggressiveness of cancer care near the end of life: is it a quality-of-care issue?. J Clin Oncol. 2008;26:3860–3866
    • View In Article
    • CrossRef
15. Matsuyama R, Reddy S, Smith TJ. Why do patients choose chemotherapy near the end of life? A review of the perspective of those facing death from cancer. J Clin Oncol. 2006;24:3490–3496
    • View In Article
    • CrossRef
16. Grunfeld EA, Maher EJ, Browne S, et al. Advanced breast cancer patients' perceptions of decision making for palliative chemotherapy. J Clin Oncol. 2006;24:1090–1098
    • View In Article
    • CrossRef
17. Grunfeld EA, Ramirez AJ, Maher EJ, et al. Chemotherapy for advanced breast cancer: what influences oncologists' decision-making?. Br J Cancer. 2001;84:1172–1178
    • View In Article
    • MEDLINE
    • CrossRef
18. Ellershaw J, Ward C. Care of the dying patient: the last hours or days of life. BMJ. 2003;326:30–34
    • View In Article
    • CrossRef
19. Sullivan AM, Lakoma MD, Matsuyama RK, et al. Diagnosing and discussing imminent death in the hospital: a secondary analysis of physician interviews. J Palliat Med. 2007;10:882–893
    • View In Article
    • CrossRef
20. Clayton JM, Butow PN, Arnold RM, Tattersall MH. Discussing end-of-life issues with terminally ill cancer patients and their carers: a qualitative study. Support Care Cancer. 2005;13:589–599
    • View In Article
    • MEDLINE
    • CrossRef