Factors Affecting the Improvement of Quality of Dying of Terminally Ill Patients with Cancer through Palliative Care: A Ten-Year Experience

Abstract

Background:

Achieving a good death is the ultimate goal of palliative medicine. Yet, very few studies have investigated factors affecting improvement in quality of dying. We therefore conducted a study to evaluate these factors in terminally ill Taiwanese cancer patients treated in a multidisciplinary palliative care unit.

Methods:

We included data from terminally ill patients with cancer admitted to the Hospice and Palliative Care Unit in the National Taiwan University Hospital from 2000 to 2009. Quality of dying was assessed by patients' multidisciplinary team at admission and after death using the Good Death Scale and the Audit Scale. We used multivariable regression to assess the association between patient factors, including gender, age, diagnosis, days of hospitalization, calendar year of admission, Good Death score at admission, and process of care scores for physical care, physician-assessed autonomy, emotional support, communication, continuity of life, and physician-reported rate of closure, with the quality of dying.

Results:

Multivariate regression analysis identified lower Good Death score at admission, lower age 40–65 years, longer unit length of stay (>7 days), higher physician-assessed autonomy, better physician-assessed emotional support, and better physician-reported rate of closure as positively related (all p<0.0001) with improvement in good death scores.

Conclusion:

In this study in a Taiwanese palliative care unit; we found that late referral to the unit and low physician-assessed autonomy were key factors negatively affecting quality of dying. Earlier truth-tellling and end-of-life care discussions between physicians and patients might improve the quality of dying in this population.

Introduction

Achieving a good death is the ultimate goal of palliative medicine.¹ Nevertheless, there are only a few well-validated instruments designed for this purpose.² Among them, Quality of Dying and Death (QODD) is the most studied and best validated.^2–4 Many published articles attempt to conceptualize and delineate the domains of a good death, but few have explored the factors affecting its achievement. Miyashita et al.⁵ found that withholding aggressive treatment and life-prolonging treatment for dying patients and use of opioids for symptom relief may be associated with achievement of a good death in Japan. The government of the United Kingdom has made it national policy to promote the good death of critically ill patients.⁶ Since achieving good death is the highest goal of palliative care, investigators in Taiwan have vigorously studied factors affecting its attainment.^7,8 Cheng et al.⁹ noted that most Taiwanese elderly are not aware of their illness. This factor compromises the achievement of patient autonomy in the terminal stage of illness. “Truth-telling” is the most important factor affecting the attainment of a good death.

Cancer has been the leading cause of death in Taiwan since 1982. More than 30,000 patients die of cancer each year and the number is increasing; however, the quality of life of those terminally ill patients with cancer is often neglected. Many patients die after tremendous suffering.¹⁰ With advances in medical technology, both Taiwan and the West have followed a similar path leading to improved palliative medicine and hospice care. The hospice movement started in 1983, and the first hospice was set up in 1990 at Mackay Memorial Hospital, a Christian-based hospital in northern Taiwan,¹¹. Many hospices were set up thereafter, mainly in private hospitals owned or sponsored by religious institutions. In 1995, the National Taiwan University Hospital (NTUH) launched the first public ward devoted to providing palliative and hospice care and aiming to integrate service with teaching and research and increase public awareness of palliative care. The Hospice and Palliative Care Unit of NTUH has served more than 2000 patients coming from all parts of Taiwan for the past 16 years, and it has been devoted to provision of a good death for Taiwanese patients. Since Hales et al.² has pointed out further research is required to understand the factors influencing the ratings of the quality of dying and death, our aim in this nonconcurrent prospective study was to examine what factors as perceived by the primary care physician affect the improvement.

Methods

Population and sampling

Our study patients were diagnosed with various terminal cancers, admitted to the Hospice and Palliative Care Unit in NTUH during the period from 2000 through 2009, and met the criteria of the study.

The inclusion criteria:

1. Stage IV malignancy refractory to available chemotherapy or any other curative treatment as determined by the original attending physician.

2. Physical symptoms requiring hospitalization, or great psychological, social, or spiritual distress.

3. Consent by the patient or family member to palliative care services, and a signed living will or do-not-esuscitate order.

The exclusion criteria:

1. Expiration within 24 hours of admission to the Hospice and Palliative Care Unit of NTUH.

2. Leukemia.

Criteria met by the primary care physician:

1. Responsibility for the patient's hospice care upon admission to hospice.

2. Basic training in palliative care for at least 4 hours as certified by the NTUH.

Our intent was to improve the validity of the questionnaire by enrolling only physicians familiar with the basic concept and principles of palliative care.

The study was approved by the Research Ethics Committee of National Taiwan University Hospital (No. 201003030R).

Data collection

For each deceased patient in hospice, the primary care physician completed a “quality of dying” evaluation form. The form had three parts: part one collected basic patient data including name, chart number, room number, gender, day of death, age, primary site of cancer, and days of hospitalization. Part two was the Good Death Scale and part three the Audit Scale (Appendix A).

The Good Death score was evaluated twice after the patient had passed away in a team conference, the first upon admission for more than 1 day to serve as a baseline and the second upon death at a usual, weekly “good death” team meeting. The team was made up of the primary care resident physician, head nurse, and study researcher (an attending physician in NTUH). The primary care physician completed the Audit Scale questionnaire after the Good Death Scale questionnaire but only after the patient's death.

Measurements

Two instruments were used in the study: the Good Death Scale and the Audit Scale. The outcome (i.e., “improvement of quality of dying”) was indicated by the difference in Good Death Scale score between admission and death.

The Good Death Scale

Derived from a modified Weisman's definition of a good death and the opinions of experienced professionals in palliative care,¹² the assessment of good death in this study had five domains: awareness that one is dying (0=complete ignorance, 3=complete awareness), acceptance of death peacefully (0=complete unacceptance, 3=complete acceptance), honoring of patient's wishes (0=no reference to the patient's wishes, 1=following the family's wishes alone, 2=following the patient's wishes alone, and 3=following the wishes of the patient and the family), death timing (0=no preparation, 1=the family alone had prepared, 2=the patient alone had prepared, and 3=both the patient and their family had prepared), and degree of physical comfort three days before death (0=a lot of suffering, 1=suffering, 2=a little suffering, and 3=no suffering).

The Audit Scale for good death services

It measures the process of patient care. The original scale was classified into 6 domains (each with 2 items). Each item was appraised on a scale of 1=extremely poor to 5=extremely good. The 6 domains and 12 items were as follows: (1) physical care: symptom control, and satisfaction of the patient and the family; (2) physician-assessed autonomy: level of autonomy and medical decision-making participation; (3) emotional support: alleviation of anxiety and resolution of depression; (4) communication: verbal support and nonverbal support; (5) continuity of life: continuity of social support and affirmation of one's past life; (6) physician-reported rate of closure: fulfillment of last wish and bereavement support.

Reliability and validity of the Good Death Scale in Taiwanese palliative care units have been well established.^8,9 Cronbach α was used to assess internal consistency of this good death measure in the present study and shown to be 0.71 for the five domains. Initially, a panel of two physicians, two nurses, one psychologist, two chaplains, and one social worker tested the entire instrument for content validity. All members of the panel were experts in palliative medicine. A content validity index was used to determine the validity of the structured questionnaire and yielded a score of 0.93. In addition, 10 volunteers (bereaved family members) filled out the questionnaire to confirm the questionnaire's face validity and ease of application.

As for the audit scale, we applied exploratory factor analysis (EFA) for testing construct validity. Bartlett's test of sphericity (BT) and the Kaiser–Meyer–Olkin (KMO) test were used to confirm that the measure was appropriate for exploratory factor analysis (BT=3373.56, KMO value=0.871, p<0.01). The draft items were analyzed using principal component factor analysis followed by orthogonal varimax rotation. The number of principal components to be extracted was determined by examining the eigen values (>1) and Cattell's Scree test. The cutoff point of factor loading in the study was set at 0.5. Finally, the number of domains were reduced to two and named “patient care” and “social well-being.” The internal consistency was demonstrated, with Cronbach α coefficient ranging from 0.84 to 0.91 in the factors and 0.93 for total items of this measure. These two factors accounted for 65.4% of the total variance of the variables.

Statistical analysis

SAS software, version 9.1.3 (SAS Institute Inc., Cary, NC) was used for data management and analysis. Two-sided p value≤0.05 was considered statistically significant.

The categorical variable data are presented as frequency (proportion) and the continuous variable data as mean (standard deviation, SD). The associations between categorical variables were analyzed by χ² test or Fisher's exact test, whereas continuous variables were compared between the posttest and pretest on the same group by paired t test or Wilcoxon rank-sum test. Means and standard deviations of the “good death” scores were compared and paired t test was used to examine the change in the Good Death Scale scores on admission and after death. Then, linear regression analysis was conducted to identify the factors predictive of a mean change in Good Death Scale score. We did a change-score analysis, i.e., ΔY=Y₁ − Y₀ was the outcome variable of our multiple linear regression model, so that the good death score at admission (i.e., Y₀) was added into the regression model as a control variable. All the univariate significant and non-significant relevant covariates (independent variables) were put on the variable list to be selected including, sex, diagnosis, days of hospitalization, calendar year of admission, Good Death Scale at admission, and categories in the Audit Scale such as physical comfort, physician-assessed autonomy, composure, communication, continuity, and physician-reported rate of closure. “Year of death” is a dummy variable to control any uninterested calendar-time associated confounding effect in our multivariate analysis. The dependent variable was the change in good death score. The significance levels for entry (SLE) and for retention or stay (SLS) were set to 0.15 or larger. Then, with the aid of clinical knowledge, the best final regression model was identified manually by reducing the significance levels to 0.05 corresponding to the chosen α level.

Results

Patient characteristics

Of the 2785 patients admitted to our Hospice and Palliative Care Unit from 2000 to 2009, a total of 2375 patients met the criteria of the study. In all, 410 patients were excluded because they died within 24 hours of admission. No physician was excluded. Table 1 shows that 51.6% of patients were older than 65 years. Males were predominant (56.3% males, 43.7% females). Almost half of the patients were admitted for less than 7 days. The top three primary tumor sites, in descending order of frequency, were hepatobiliary (22.2%), gastrointestinal (21.2%), and respiratory (19.8%). The average number of days of admission was 10.8.

Table 1.

Demographic Data of Patients Admitted to Hospice and Palliative Care Unit, National Taiwan University Hospital from 2000–2009

Variable	n	%
Age (years)	Mean (SD)=63.81±15.13
<12	5	0.2
12–18	5	0.2
19–35	59	2.5
36–65	1078	45.5
>65	1225	51.6
Gender
Male	1338	56.3
Female	1037	43.7
Days of admission (days)	Mean (SD)=10.79±9.49
1–7	1173	49.4
8–30	1105	46.5
>31	96	4
Primary sites of tumor
Hepatobiliary	525	22.2
Gastrointestinal	501	21.2
Respiratory	467	19.8
Head and neck	218	9.2
Gynecological	129	5.5
Breast	108	4.6
Pancreas	105	4.5
Hematologic	33	1.4
Musculoskeletal	25	1.1
Unknown	216	9.2
Others	31	1.3

SD, standard deviation.

Good Death Scale

From Table 2 it is apparent that all five domains of the Good Death Scale were improved after palliative care in hospice (p<0.0001) and the mean increase in total score was 2.6 on a scale of 15.

Table 2.

Difference in Various Domains of Good Death Score between Admission and Prior to Death

Domains	At admission (mean)	At death (mean)	Difference (mean)	SD	t	p
Awareness	2.10	2.68	0.58	0.66	42.64	<0.0001
Acceptance	2.21	2.71	0.50	0.61	39.29	<0.0001
Propriety	2.20	2.79	0.21	0.60	17.38	<0.0001
Timeliness	2.09	2.78	0.59	0.91	31.50	<0.0001
Comfort	2.09	2.82	0.72	0.64	54.87	<0.0001
Total score	11.21	13.80	2.60	2.35	53.13	<0.0001

SD, standard deviation.

The Audit Scale

Table 3 shows the distribution of audit scale scores. Most domains were rated “good” and “extremely good.”

Table 3.

Results of Audit Scale at Death from Patients Admitted to Hospice and Palliative Care Unit, National Taiwan University Hospital from 2000–2009

	Extremely poor	Poor	Fair	Good	Extremely good
Categories of audit scale	1	2	3	4	5
Physical care
Symptom control	0	5 (0.002%)	153 (6.4%)	1324 (55.7%)	893 (37.6%)
Patient and family satisfaction	0	2 (0.0008%)	88 (3.7%)	1174 (49.4%)	1111 (46.8%)
Physician-assessed autonomy
Level of autonomy	0	8 (0.34%)	132 (5.6%)	1127 (47.7%)	1095 (46.4%)
Medical decision-making participation	1 (0.0004%)	7 (0.3%)	121 ((5.1%)	1106 (46.8%)	1127 (47.7%)
Emotional support
Alleviation of anxiety	0	10 (0.4%)	145 (6.1%)	1279 (53.9%)	938 (39.5%)
Resolution of depression	0	12 (0.5%)	148 (6.2%)	1277 (53.8%)	935 (39.4%)
Communication
Verbal support	0	0	50 (2.1%)	1091 (46%)	1230 (51.9%)
Nonverbal support	0	1 (0.0004%)	36 (1.5%)	1044 (44%)	1290 (54.4%)
Continuity of life
Continuity of social support	0	10 (0.004%)	106 (4.5%)	1007 (42.5%)	1244 (52.6%)
Affirmation of one's past life	2 (0.0008%)	6 (0.25%)	107 (4.5%)	1095 (46.3%)	1157 (48.9%)
Physician-reported rate of closure
Fulfillment of last wish	1 (0.0004%)	11 (0.5%)	176 (7.4%)	1132 (47.7%)	1053 (44.4%)
Bereavement support	1 (0.0004%)	2 (0.08%)	66 (2.8%)	999 (42.1%)	1305 (55%)

Predictors of improvement of a good death

Univariate analysis showed that younger age (<40 years, p<0.0001), gastrointestinal tumors (p<0.0001), unknown primary site of tumors (p=0.0184), days of admission (p<0.0001), admission days>7 (p<0.0001), Good Death Scale score at admission, higher physician-assessed autonomy from the Audit Scale (p=0.0153), and better emotional support from the audit scale (p=0.0145; Table 4) were significantly related to higher good death score. Aside from GDS score at admission (which was negatively related), multivariate analysis identified age between 40–65 years (p<0.0001), number of admission days more than 7 (p<0.0001), higher physician-assessed autonomy (p<0.0001) from the audit scale, better emotional support (p<0.0001), and physician-reported rate of closure (p<0.0001) as factors affecting the improvement of a good death. In evaluating the effects of categorical variables, the categories of a categorical variable (e.g., age) without the associated beta values were combined as the joint reference group for that categorical variable to gain efficiency in statistical estimation. Any discrepancy between the results of univariate analysis and multivariate analysis was likely caused by the confounding effects of the uncontrolled covariates in the univariate analysis.

Table 4.

Factors Associated with the Improvement of Good Death Scale Score for Patients Admitted to Hospice and Palliative Care Unit, National Taiwan University Hospital from 2000–2009

	Univariate analysis			Multivariate analysis
Predictors	β	95% CI	p value	β	95% CI	p value
Intercept				4.1399
Age	−0.0059	−0.012–0.00053	0.0721
<40	2.6240	2.21–3.04	<0.0001^*
40>, <65	0.0689	−0.37–0.51	0.7582	0.2461	0.06302	<0.0001^*
>65	−0.094	−0.53–0.34	0.6709
Gender (male)	−0.0358	−0.23–0.16	0.7169
Primary site of malignancy	−0.0108	−0.04–0.02	0.5001
Gastrointestinal	2.7608	2.55–2.97	<0.0001^*
Respiratory	−0.2279	−0.53–0.07	0.1368
Hepatobiliary	−0.1436	−0.44–0.15	0.3344
Breast	−0.4368	−0.93–0.05	0.0806
Pancreas	−0.1686	−0.67–0.33	0.5082
Gynecological	−0.1848	−0.65–0.28	0.4327
Head & neck	−0.3101	−0.69–0.069	0.1082
Musculoskeletal	−0.1608	−1.11–0.78	0.7384
Unknown	−1.0108	−1.85–-0.17	0.0184^**
Hematologic	0.0323	−0.85–0.91	0.9427
Other	−0.0535	−0.44–0.33	0.7845
Year of death
2000	1.2212	0.8–1.65	<0.0001^*
2001	1.0123	0.48–1.54	0.0002^*
2002	0.9915	0.48–1.51	0.0002^*
2003	1.0267	0.51–1.54	<0.0001^*
2004	2.251	1.74–2.76	<0.0001^*	0.3976	0.10122	<0.0001^*
2005	1.7073	1.20–2.22	<0.0001^*	0.4143	0.10005	<0.0001^*
2006	1.7101	1.20–2.22	<0.0001^*
2007	1.2607	0.75–1.77	<0.0001^*
2008	1.8783	1.36–2.40	<0.0001^*	0.3445	0.10463	0.0010^***
2009	1.1362	0.63–1.64	<0.0001^*
Day of admission	0.025	0.015–0.03	<0.0001^*
>7 days	0.023	0.014–0.03	<0.0001^*	−0.6721	0.00285	<0.0001^*
Good Death Scale at admission	−0.596	<2e–16^*	−0.6721	0.0117	<2e–16^*
Physical Care (Audit Scale)	−0.0477	−0.14–0.04	0.2924
Symptom control	−0.1329	−0.30–0.03	0.1055
Patient and family Satisfaction	−0.026	−0.19–0.14	0.7624
Physician-assessed autonomy (Audit Scale)	−0.1003	−0.18–-0.02	0.0153^**	0.1522	0.03439	<0.0001^*
Honoring of autonomy	−0.2151	−0.37–-0.06	0.0074^***
Medical decision-making Participation	−0.1643	−0.32–-0.006	0.0417^**
Emotional support (Audit Scale)	−0.0685	−0.15–0.01	0.0981	0.1666	0.03521	<0.0001^*
Alleviation of anxiety	−0.0608	−0.22–0.10	0.4498
Resolution of depression	−0.1956	−0.35–-0.039	0.0145^**
Communication (Audit Scale)	0.0626	−0.03–0.15	0.1819
Verbal support	0.0674	−0.11–0.24	0.4564
Nonverbal support	0.1718	−0.01–0.35	0.0622
Continuity of life (Audit Scale)	0.0424	−0.04–0.13	0.3309
Continuity of social support	0.1367	−0.02–0.3	0.0926
Affirmation of one's past life	0.0108	−0.15–0.17	0.8943
Physician reported rate of closure (Audit Scale)	0.0001	−0.089–0.089	0.998	0.3246	0.03873	<0.0001^*
Fulfillment of last wishes	−0.1207	−0.27–0.03	0.1126
Bereavement support	0.1582	−0.11–0.33	0.0689

p<0.001.

p<0.05.

***

p<0.01.

CI, confidence interval.

Discussion

Using physician-reported outcomes, our study is the first non-concurrent prospective investigation of factors affecting the improvement of quality of dying of terminally ill patients with cancer in a palliative care unit in Asia. Age 40–65 years, days of admission more than 7, low Good Death Scale score at admission, higher physician-assessed autonomy, better emotional support, and better closure were identified as significant factors affecting the improvement of good death.

We took older patients (>65 years) as the reference group in our multivariate analysis. Compared to the older patients (>65 years), the patients with age younger than 40 did not have a statistically significantly different mean value of the outcome variable, so that these two age groups of patients were combined as the joint reference group to gain efficiency in statistical estimation (Table 4).

In all, 49.4% of patients passed away within a week of entering hospice. The relatively short stay may be attributed to late referral from the original unit and limited resources for palliative care. From multiple regression analysis, it can be concluded that quality of dying, as perceived by the medical team, is better when the hospice stay is more than a week (p<0.0001).

Late referral to hospice is apparently a common problem in Asian countries such as Japan¹³ and Korea¹⁴ as well. In Japan, the reasons for late referral are the family's misconception about hospice care, the inadequacy of end-of-life discussions between physician and family, and the inadequacy of the family's preparation for deterioration of patient's condition.¹³ The passing of the Cancer Control Act in 2006 has increased the number of palliative care teams actively involved in facilitating the family's access to palliative care services.¹⁵ In Korea, the reasons for late referral are mainly socioeconomic and medical factors.¹⁴

There are many reasons for late referral in Taiwan; one reason is easy access to medical care. A retrospective study investigating trends in end-of-life care in Taiwan (2000–2006) revealed the tendency toward more aggressive cancer care.¹⁶ Tang et al.¹⁷ concluded that aggressive end-of-life care significantly affects the decision of very terminally ill patients to choose hospice care in Taiwan. Patients with cancer in Taiwan utilize the medical resources covered by National Health Insurance for treatment until the last minute of their lives.

Although access to medical care is easy in Taiwan, access to palliative care is impeded by many complex issues such as truth-tellling, which can delay referral to hospice. The Taiwanese has a tendency to conceal the “bad news” from the patient in afraid that the patient could not take it psychologically.⁹ Truth-telling and end-of-life care discussions between physicians and patients should be encouraged as early as possible to ease the transition from curative to palliative care in Taiwan.¹⁸ Late referral to the hospice will inevitably impede the process of dying preparation and therefore cause lower Good Death Scale score at death.

“Physician-assessed autonomy” from the Audit Scale significantly affected quality of dying of terminally ill patients with cancer in our study. It is classified into two categories; one is related to honoring of autonomy (the extent to which a patient is allowed to decide what is best for him/her, such as whether to eat). The other is related to the patient's ability to participate in medical-decision making (i.e., health care decisions, such as do not resuscitate, use of narcotics). All assessments were retrospective and proxy in nature and based on data from medical charts or nurses' records. If the patient always had a high degree of autonomy, the rating would be “extremely good.” Conversely, if the patient never had autonomy, the rating would be “extremely poor.” Unlike western countries where patient autonomy is generally respected, in Asia, it is common practice to let the family decide for the patient especially to make the medical decisions. In Japan, Morita et al.¹⁹ noted that loss of autonomy was one of the three principal components of existential suffering of Japanese terminally ill patients with cancer. Likewise, a study by Cheng et al.⁹ reported that truth-telling compromises autonomy in elderly patients with terminal cancer and consequently affects their good death scores. Recently, a Korean study by Mo et al.²⁰ questioned the value of patient autonomy in terminally ill patients with cancer. Nevertheless, we found it is cross-sectional and small sample size (n=93) in study design and up to 78.5% of the Korean patients had awareness of their terminal illness, which is quite different from the situation in Taiwan.

There were several limitations in the study. All the data were gathered and evaluated by primary care physicians after the patients had passed away due to the nature of the study. Proxy data collection is prevalent in palliative medicine. The issue is important since the validity and reliability of proxy data have been challenged repeatedly. However, in our study all data were collected in a standardized way under supervision to reduce the bias as much as possible. Since the data were mainly gathered by questionnaires retrospectively, the study was prone to recall bias. To reduce recall bias, we required that the questionnaires be completed no later than 1 week after the patient had passed away.

Conclusions

In this 10-year nonconcurrent prospective study, data from 2375 terminal cancer patients coming from all parts of Taiwan were analyzed and factors affecting the improvement of quality of dying were investigated. We concluded higher physician-assessed autonomy, better emotional support, and better physician-reported rate of closure from the audit scale in patients with age 40–65 years, more than 7 days of hospice stay (days of admission>7), and low Good Death Scale score at admission as factors significantly associated with good death outcome. More resources should be devoted to ensuring early referral to hospice and palliative care and patient autonomy. Truth-tellling and end-of-life care discussions between physicians and patients should be initiated as early as possible to prepare the patients for better quality of dying. A nationwide, multicenter survey of “good death” in Taiwanese people should be conducted in the future.

Footnotes

Acknowledgments

The authors would like to thank the faculties of the Department of Family Medicine, National Taiwan University Hospital, and also to C.C Chang and Y.H. Li for their assistance in preparing this manuscript.

Author Disclosure Statement

No competing financial interests exist.

Appendix A. Quality of Dying Evaluation Form in the Hospice and Palliative Care Unit at National Taiwan University Hospital

References

Storey

. Goals of hospice care. Tex Med, 1990; 86:50–54.

Hales

, Zimmerman

, Rodin

. Review: The quality of dying and death: A systematic review of measures. Palliat Med, 2010; 24:127–144.

Curtis

, Patrick

, Engelberg

, Norris

, Asp

, Byock

. A measure of the quality of dying and death. Initial validation using after-death interviews with family members. J Pain Symptom Manage, 2002; 24:17–31.

Downy

, Curtis

, Lafferty

, Herting

, Engelberg

. The Quality of Dying and Death Questionnaire (QODD): Empirical domains and theoretical perspectives. J Pain Symptom Manage, 2010; 39:9–22.

Miyashita

, Morita

, Sato

, Hirai

, Shima

, Uchitomi

. Factors contributing to evaluation of a good death from the bereaved family member's perspective. Psychooncology, 2008; 17:612–620.

Coombs

, Long

. Managing a good death in critical care: Can health policy help? Nur Crit Care, 2008; 13:208–214.

Tsai

, Wu

, Chiu

, Hu

, Chen

. Fear of death and good death among the young and elderly with terminal cancers in Taiwan. J Pain Symptom Manage, 2005; 29:344–351.

Yao

, Hu

, Lai

, Cheng

, Chen

, Chiu

. Does dying at home influence the good death of terminal cancer patients? J Pain Symptom Mange, 2007; 34:497–504.

Cheng

, Hu

, Liu

, Yao

, Chen

, Chiu

. Good death study of elderly patients with terminal cancer in Taiwan. Palliat Med, 2008; 22:626–632.

10.

Sepulveda

. New Guide on Palliative Care Services for People Living with Advanced Cancer. Geneva: World Health Organization, 2007. www.who.int/mediacentre/news/notes/2007/np31/en/. 2012 June> 11.

11.

Lai

, Su

. Palliative medicine and the hospice movement in Taiwan. Support Care Cancer, 1997; 5:348–350.

12.

Weisman

. Appropriate death and the hospice program. Hosp J, 1988; 4:65–77.

13.

Morita

, Akechi

, Ikenaga

, Kizawa

, Kohara

, Mukaiyama

et al. Late referrals to specialized palliative care service in Japan. J Clin Oncol, 2005; 23:2637–2644.

14.

Baek

, Shin

, Choi

, Kang

, Mo

, Kim

et al. Late referral to palliative care services in Korea. J Pain Symptom Manage, 2011; 41:692–699.

15.

Morita

, Miyashita

, Tsuneto

, Sato

, Shima

. Late referrals to palliative care units in Japan: Nationwide follow-up survey and effects of palliative care team involvement after the Cancer Control Act. J Pain Symptom Manage, 2009; 38:191–196.

16.

Tang

, Wu

, Hung

, Huang

, Chen

, Liu

. Trends in quality of end-of-life care for Taiwanese cancer patients who died in 2000–2006. Ann Oncol, 2009; 20:343–348.

17.

Tang

, Huang

, Liu

, Wang

, Rau

, Chen

. Aggressive end-of-life care significantly influenced propensity for hospice enrollment within the last three days of life for Taiwanese cancer decedents. Pain Symptom Manage(in press).

18.

Tang

, Huang

, Liu

, Wang

, Chen

. A population-based study on the determinants of hospice utilization in the last year of life for Taiwanese cancer decedents, 2001–2006. Psychooncology, 2010; 19:1213–1220.

19.

Morita

, Tsunoda

, Inoue

, Chihara

. An exploratory factor analysis of existential suffering in Japanese terminally ill cancer patients. Psychooncology, 2000; 9:164–168.

20.

, Shin

, Woo

, Choi

, Kang

, Baik

et al. Is patient autonomy a critical determinant of quality of life in Korea? End-of-life decision making from the perspective of the patient. Palliat Med(in press).