Use of chemotherapy at the end of life among Taiwanese cancer decedents, 2001–2006

Abstract

Introduction. The availability of new chemotherapeutic agents has lengthened the treatment timeline for advanced cancers and increases the likelihood of receiving chemotherapy near death. Use of chemotherapy near the end of life may not benefit cancer patients. However, no population-based study has examined the determinants for continuing chemotherapy at the end of life for all ages and cancer groups as well as for a whole country. This population-based study assessed the association between continuation of chemotherapy in the last month of life and patient demographics, disease characteristics, primary physician's specialty, hospital characteristics, and healthcare resource availability at the hospital and regional levels. Materials and methods. Retrospective population-based cohort study using administrative data among 204 850 Taiwanese cancer decedents in 2001–2006. Results. Rates of continued chemotherapy in the last month of life for each study year were 17.5%, 17.4%, 17.3%, 19.0%, 20.0%, and 21.0%, respectively and have remained steady since 2001. Taiwanese cancer patients had greater odds for continuation of chemotherapy in the last month of life if they were male [adjusted odds ratio (AOR) 1.19, 95% confidence interval (CI) 1.13–1.25], younger, single [1.21 (1.09–1.35)], had lower comorbidity levels, were diagnosed with hematologic malignancies [1.90 (1.09–1.35)] and breast cancer [1.24 (1.08–1.43)], had metastatic disease [1.36 (1.27–1.46)], and survived <  1 year but longer than two months post-diagnosis. The odds for continued chemotherapy in patients’ last month was significantly increased by being cared for by a medical oncologist [3.49 (3.04–3.99)] or in a teaching hospital [1.39 (1.11–1.74)] and with the highest intensity of total inpatient hospital beds [1.63 (0.99–2.68)], but was not influenced by regional healthcare resources (total hospital and hospice beds). Conclusion. The relative risk for continuation of chemotherapy in the last month of life was determined by patient demographics and disease characteristics, physician specialty, and healthcare resources at the primary hospital level.

The availability of new chemotherapeutic agents has lengthened the treatment timeline for advanced cancers and increases the likelihood of receiving chemotherapy near death [1–4]. However, the use of chemotherapy near the end of life (EOL) may not benefit cancer patients, as evident by its precipitating emergency room visits [5], increasing intensive care unit care [4,5], precluding early hospice referral [1], leading to highly frequent deaths in a hospital [1,6], elevating anxiety and depression [7], and decreasing satisfaction with care [2]. Therefore, overly aggressive chemotherapy use at EOL has been recognized as representing poor quality cancer care [2,8].

As more cancer patients are being offered chemotherapy close to death [1–4], information on how chemotherapy is administered to which patients and the impact of continuing chemotherapy close to death on patient quality of life (QOL) and EOL care received will enable cancer patients, families, and clinicians to make more informed decisions about whether to start or continue chemotherapy in this difficult period. Better understanding the sources of variation in the use of chemotherapy at EOL not explained by patients’ disease characteristics may facilitate tailoring therapy to those patients most likely to benefit. However, factors predisposing cancer patients to receive chemotherapy in the last months of life have only been explored in the elderly [1,2,4,9], for selected cancers [1,3,4,7,9,10], at one to two centers [6,7,11–13] or in limited geographic areas [3,10,14,15]. No population-based study has been conducted on the determinants for continuing chemotherapy at EOL for all ages and cancer groups as well as for a whole country. Therefore, the purpose of this population-based study was to assess the association between continuation of chemotherapy in the last month of life and patient demographics, disease characteristics, primary physician's specialty, hospital characteristics, and availability of healthcare resources at the hospital and regional levels.

Materials and methods

Study design and sample

A retrospective cohort design was used by linking individual patient-level data with encrypted personal identification numbers from computerized data of the National Register of Deaths Database (NRDD), Cancer Registration System (CRS) database, National Health Insurance (NHI) claims database, and the Database of Medical Care Institutions Status (DMCIS). These databases were monitored for completeness and accuracy by Taiwan's Department of Health. Taiwan's NHI has the unique characteristics of universal insurance coverage, comprehensive services provided, a single-payer system with the government as sole insurer, and coverage of 99.0% of Taiwan's population [16]. Patients have free access to any healthcare system and provider they choose. Healthcare systems are reimbursed for services provided, and co-payment is waived for patients with malignancy. The CRS included 97.3% of incident cancer cases, with 99.6% completeness and 87.3–91.4% accuracy [17]. The DMCIS provided information on hospital characteristics and healthcare resources for each hospital and region.

Measures

The outcome variable.Due to data availability, we examined use of chemotherapy in the last month of life rather than within 14 days of death as originally developed by Earle et al. [8]. Use of chemotherapy was identified from both inpatient and outpatient NHI claims, which specified the codes for all routes of chemotherapy administration but oral.

Independent variables.The outcome variable was hypothesized to be associated with the following four groups of variables based on the literature [2,3,15]. A year-indicator variable was also included to examine national trends in EOL care in Taiwan.

Patient demographics and disease characteristics. Differences in outcome variables were examined across gender and four age categories (< 65, 65–74, 75–84, ≥85years). Marital status at time of death was categorized as married, single, divorced or separated, and widowed.

Disease characteristics included comorbidities, cancer diagnosis, and post-diagnosis survival time. Comorbidities were identified from ICD-9 codes for the primary and secondary diagnoses, excluding cancer-related codes, in NHI claims for both inpatients and outpatients in months 1 to 12 before death. These ICD-9 codes were used to calculate the Deyo-Charlson comorbidity index [18], which was modified from the Charlson comorbidity index [19] by replacing data abstracted from medical record review to ICD-9-CM codes.

Diagnosis and date of diagnosis were identified from the CRS, and metastatic status was identified by ICD-9 codes in inpatient and outpatient claims data that included codes 196–199. Survival time was calculated as the interval (in months) between dates of diagnosis and death and was further categorized into 1–2, 3–6, 7–12, 13–24, and ≥25 months.

Primary physician's specialty. In Taiwan, chemotherapy can be prescribed by physicians other than oncologists, especially for newly diagnosed cancer patients. The patient's primary physician for EOL care was defined as the first physician seen in the patient's last month of life. The primary physician's specialty was retrieved from the code for reported physician specialty in NHI claims and was dichotomized into medical oncologist (including hematologists but excluding radiation oncologists, surgical oncologists, and gynecological oncologists) and other.

Characteristics and healthcare resources of the primary hospital. The patient's primary hospital was identified as the hospital where his/her primary physician served. Characteristics and healthcare resources of the primary hospital included teaching status, availability of an inpatient hospice unit, and bed capacity—all obtained from the DMCIS. Clinical consultant teams with palliative specialty credentials were not available in Taiwan in 2001–2005. Furthermore, hospice home care was limited. Therefore, “availability of an inpatient hospice unit” represented the hospice resource of the primary hospital.

Healthcare resources in the primary hospital's region. Patients were assigned to a region that included the county/city for their primary hospital, and in turn to a quartile for hospital beds. Regional hospice beds per 10 000 population were calculated for each region that included the patient's primary hospital. Healthcare resources in the primary hospital's region were ascertained from the DMCIS.

Statistical analysis

Multivariate logistic regression by the generalized estimating equation method (GEE) [20] was used to examine the impact of each explanatory variable on use of chemotherapy in the last month of life. The GEE model is characterized by robust standard errors accounting for correlation in the error term due to clustering of individuals in the same hospital and by its simultaneous adjustment for all other factors in the model.

Results

The NRDD identified 210 976 cancer deaths (including all ages and disease groups of cancer patients) in 2001–2006. Among these cancer decedents, 6126 were excluded from this study because they did not submit any NHI claims for inpatient or outpatient care in the last month of life, leaving a study sample of 204 850 patients who had received medical care in the last month of life (for subject characteristics, see Table I).

Table I. Table I. Determinants of using chemotherapy in the last month of life for Taiwanese cancer decedents, 2001–2006.

	Decedents
Parameter	n	%	Chemo-therapy (%)	Adjusted odds ratio	95% Confidence limits		p
Year ^a				1.02	0.99	1.04	0.22
Gender
Male	131 614	64.3	17.0	1.19	1.13	1.25	< 0.0001
Female	73 236	35.8	17.1	Ref ^b
Age (years)
< 65	83 072	40.6	25.8	Ref
65–74	55 400	27.0	15.1	0.59	0.57	0.62	< 0.0001
75–84	52 073	25.4	8.9	0.35	0.32	0.37	< 0.0001
≥ 85	14 305	7.0	3.3	0.15	0.12	0.17	< 0.0001
Marital status
Married	141 157	68.9	18.2	Ref
Single	14 591	7.1	23.5	1.21	1.09	1.35	0.0005
Divorced/separated	6843	3.3	21.0	0.95	0.87	1.03	0.19
Widowed	38 674	18.9	9.7	0.89	0.84	0.93	< 0.0001
Comorbidities ^c
0 or 1	96 281	47.0	22.4	Ref
2	39 421	19.2	16.0	0.89	0.85	0.93	< 0.0001
3	23 526	11.5	13.5	0.78	0.75	0.82	< 0.0001
≥ 4	45 622	22.3	8.3	0.61	0.57	0.66	< 0.0001
Cancer group
Lung	40 680	19.9	20.1	Ref
Hematological	11 444	5.6	40.9	1.90	1.55	2.34	< 0.0001
Breast	7701	3.8	28.9	1.24	1.08	1.43	0.003
Liver	40 431	19.7	5.9	0.29	0.24	0.37	< 0.0001
Brain	2728	1.3	7.6	0.30	0.23	0.41	< 0.0001
Prostate	4617	2.3	4.2	0.30	0.23	0.38	< 0.0001
Pancreatic/bile duct	11 982	5.9	14.0	0.55	0.48	0.64	< 0.0001
Esophageal	6661	3.3	19.9	0.59	0.50	0.70	< 0.0001
Head and neck	17 040	8.3	23.9	0.79	0.70	0.90	0.0004
Gastric	13 993	6.8	16.9	0.81	0.69	0.96	0.01
Colon-rectum	21 998	10.7	15.7	0.82	0.73	0.91	0.0003
Other	25 575	12.5	16.7	0.66	0.60	0.73	< 0.0001
Metastasis
No	77 351	37.8	11.5	Ref
Yes	127 499	62.2	20.4	1.36	1.27	1.46	< 0.0001
Post-diagnosis survival (months)
1–2	29 970	16.1	12.6	Ref
3–6	36 763	19.8	21.3	1.62	1.48	1.77	< 0.0001
7–12	33 154	17.9	21.5	1.23	1.09	1.39	0.0008
13–24	34 003	18.3	19.1	0.94	0.83	1.07	0.38
≥ 25	51 890	27.9	15.0	0.72	0.63	0.83	< 0.0001
Primary physician's specialty
Other	159 249	77.7	11.1	Ref
Oncologist	45 601	22.3	37.7	3.49	3.04	3.99	< 0.0001
Hospital teaching status
No	19 555	9.6	7.2	Ref
Yes	184 882	90.4	18.1	1.39	1.11	1.74	0.004
Hospital beds
<Q1	50 982	25.0	8.7	Ref
Q1–Q3	100 033	49.0	17.7	1.18	0.84	1.67	0.34
>Q3	53 191	26.1	23.9	1.63	0.99	2.68	0.05
Hospital availability of hospice service
No	120 730	58.9	15.3	Ref
Yes	84 120	41.1	19.4	0.88	0.58	1.35	0.56
Regional hospice beds/10 000 population				0.84	0.52	1.37	0.49
Regional hospital beds
<Q1	50 861	24.9	13.8	Ref
Q1–Q3	98 132	48.0	18.1	1.12	0.96	1.32	0.15
>Q3	55 385	27.1	18.2	1.11	0.91	1.35	0.32

Q, quartile

^a For each successive year

^b Reference group

^c Comorbidities were assessed by the Deyo-Charlson comorbidity index

Rates for continued chemotherapy in the last month of life for each study year were 17.5%, 17.4%, 17.3%, 19.0%, 20.0%, and 21.0%, respectively. The results from GEE analysis indicated no significant trend toward more chemotherapy use at EOL from 2001 to 2006 (p =  0.22, Table I).

The likelihood of chemotherapy being continued in the last month of life was influenced by several patient demographic and disease characteristics. Males were more likely to receive chemotherapy at EOL [adjusted odds ratio (AOR) 1.19, 95% confidence interval (CI) 1.13–1.25] (Table I). Continuation of chemotherapy in the last month of life decreased sharply with age and had age gradients (as shown by no overlap in 95% CIs of AORs). Compared to married patients, single patients were significantly more likely to receive chemotherapy in their last month of life, but widowed patients used significantly less. Comorbidity level progressively decreased the odds of using chemotherapy in the last month of life. Compared to patients with lung cancer (Taiwan's leading cause of cancer death), patients with hematological malignancies and breast cancer were significantly more likely to receive chemotherapy [AOR (95% CI) 1.90 (1.55–2.34) and 1.24 (1.08–1.43), respectively]. In contrast, patients with all other cancers, especially hepatoma, brain tumor, as well as prostate and pancreatic/bile duct cancers were less likely to continue chemotherapy in the last month of life. Cancer patients with a distant-metastatic disease were more likely to use chemotherapy before death [AOR (95% CI) 1.36 (1.27–1.46)]. Cancer patients who survived up to one year after diagnosis had higher odds of using chemotherapy in the last month of life than patients who survived only one to two months, but these patients had the same likelihood as those who survived 13–24 months after diagnosis. Surviving longer than two years post-diagnosis significantly decreased the odds of continuing chemotherapy before death (AOR 0.72, 95% CI 0.63–0.83).

Cancer patients cared for by a medical oncologist as their primary physician had 3.49 times (95% CI 3.04–3.99) higher odds than patients cared for by other physician specialists of using chemotherapy in their last month (Table I). The likelihood of continued chemotherapy in the last month of life was not influenced by the primary hospital's availability of an inpatient hospice unit. However, cancer patients were 1.39 times more likely to continue chemotherapy in their last month if they received care in a teaching hospital (95% CI 1.11–1.74) than in a non-teaching hospital. The odds of continuing chemotherapy in the last month of life increased with increasing quartile of total hospital beds in the primary hospital, reaching a significant level when the number of total hospital beds was greater than the third quartile [AOR (95% CI) 1.63 (0.99–2.68)]. In contrast, the use of chemotherapy at EOL was not associated with a higher density of total hospital beds or inpatient hospice beds in the region of the decedent's primary hospital.

Discussion

Taiwanese cancer decedents from 2001 to 2006 continued chemotherapy in the last month of life at rates of 17.3–21.0%. Receiving chemotherapy within two weeks of death has been reported as 2.4–6.4% in Canada [10,14,21], 4.3% in Japan [22], 4.6–20% in the US [1–3,9], and 21% in Portugal [11]. More closely related to our study's time frame, continuation of chemotherapy in the last month of life has been documented as 9–43% in the US [7,15], 5.1–37% in Portugal [6,11], 30.9% in Korea [23], 22.7% in Italy [12], 19.7% in Finland [24], and 18% in Australia [13]. The proportions of Taiwanese cancer patients who continued chemotherapy close to death are at the low end of worldwide statistics. Furthermore, in contrast to the increasing trend to continue chemotherapy at EOL observed in the US-based studies [1–4], rates for chemotherapy in the last month of life for Taiwanese cancer decedents have remained steady in 2001–2006.

A higher relative risk for continuing chemotherapy in the last month of life was determined by several characteristics of Taiwanese patients, as reported, i.e. male gender [2,14], younger age [2,6,11,15,23], and lower comorbidity levels [1,6]. Of note, increasing age and comorbidity levels had incremental effects on decreasing the likelihood to continue chemotherapy at EOL. Our finding that male cancer patients were more likely to use chemotherapy in their last month of life may derive from gender differences in preferences for EOL care and participation in discussing EOL care. Men have been reported to have a stronger preference than women for life-sustaining treatment [25], whereas women were more likely than men to participate in discussions about EOL care [26]. Therefore, women are more likely to use hospice care at EOL in contrast to more men using inpatient services [27], which may include chemotherapy at EOL. Nevertheless, whether terminally ill male cancer patients really prefer more chemotherapy at EOL warrants further investigation to directly examine the extent to which continuation of chemotherapy in male cancer patients’ last month of life reflects their preference for aggressive anti-cancer treatment at EOL.

Married patients were documented to be more likely than unmarried patients to receive chemotherapy at EOL [2,23]. However, our results show that unmarried cancer patients were not uniform in their likelihood of receiving chemotherapy in the last month of life; singles were more likely and widowed patients were less likely than married patients to use chemotherapy at EOL. Patients have been shown to obtain tangible clinical benefits from marriage, including using higher quality hospitals [28], greater survival [29], better QOL [29] and pain control [28], and dying at home [30]. Spouses may exert a protective role as higher-order decision-makers, which may be lacking in single cancer patients, to avoid unnecessary and potentially futile chemotherapy. On the other hand, widowed cancer patients were found to have a lower likelihood of receiving chemotherapy [2,23,29], which has been attributed to advanced age and lack of day-to-day practical assistance through their widowhood [28].

Continuation of chemotherapy in Taiwanese cancer patients’ last month of life varied as a function of cancer type, in contrast to reports indicating that cancer chemosensitivity did not influence whether chemotherapy was received at EOL by dying patients from the USA [15], Italy [12], and Australia [13]. In line with the literature, Taiwanese patients had a greater relative risk of using chemotherapy in their last month of life if they had hematological malignancies [2,15] and breast cancer [2,6,11]. Moreover, patients with lung cancer had a higher probability of using chemotherapy in their last month of life than patients with cancer generally considered unresponsive to chemotherapy, such as hepatocellular, pancreatic, or bile duct cancers [23], and tumors of the central nervous system, head and neck, and prostate [14]. Our results also indicate that patients with lung cancer were more aggressively treated with chemotherapy in their last month of life than patients with colon-rectal, gastric, and esophageal cancers. This phenomenon may be related to the limited number of available chemotherapeutic agents reimbursed by the NHI during 2001–2006 for colon-rectal, gastric, and esophageal cancers. However, rates of continued chemotherapy at EOL for these cancers have not been explored in other studies and warrant further investigation.

Metastatic status and post-diagnosis survival time were significant determinants of whether Taiwanese cancer patients would continue chemotherapy at EOL. These results are not totally consistent with the literature. Metastatic diseases have been positively [6] and negatively [2] associated with use of chemotherapy at EOL. The proportion of patients still receiving chemotherapy at EOL was reported as decreasing monotonically with increasing survival [2] or not changing [11]. Research is needed to illustrate the impact of cancer metastatic status and post-diagnosis survival time on predisposing patients to continue chemotherapy at EOL.

Taiwanese cancer patients cared for by medical oncologists were 3.49 (95% CI 3.04–3.99) times more likely than patients cared for by other physician specialists to continue chemotherapy in their last month of life, in line with reports from Western countries [1–3]. Medical oncologists had more positive attitudes toward and better knowledge of the role of palliative chemotherapy in metastatic cancer than other specialist groups [31]. Therefore, they may administrate chemotherapy more commonly for cancer patients at EOL.

Alternatively, the continuation of chemotherapy close to death may reflect medical oncologists’ respect for a decision made by cancer patients and their family caregivers. Cancer patients are willing to endure major adverse effects for a small and often non-significant survival gain to avoid death at any cost [32]. They request that patient preferences for treatment should direct use of chemotherapy at EOL [33]. Having long-term contact with cancer patients, medical oncologists may be the most aware of how desperate these patients are for small survival benefits and empathetic with the dilemma experienced by these patients and families while they struggle in adjusting to the prognosis. Thus, medical oncologists may agree to provide chemotherapy to give them more time in treatment so they can adjust to the reality that more treatment is not curing their disease.

Taiwanese cancer patients receiving care in hospitals with teaching responsibilities and a higher density of inpatient hospital beds were significantly more likely than their counterparts to continue chemotherapy in their last month of life. However, conflicting results have been reported on teaching hospital status [2,3] and total hospital beds [3,23] in determining the relative risk for continuing chemotherapy at EOL. We found that availability of hospice care was not associated with decreased odds of received chemotherapy, in contrast to previous reports [3,14,23]. Further investigation is warranted on the influence of healthcare resources at the primary hospital level on continuation of chemotherapy at EOL for terminally ill cancer patients.

Regional healthcare resources (total hospital and hospice beds) did not significantly influence Taiwanese cancer patients’ relative risk for continuing chemotherapy in their last month of life. The discrepancy between our results and those in the literature [2] might be due to differences in defining “region”. Defining a region by location of the patient's primary hospital rather than by his/her residence may result in that hospital's beds overriding the influence of regional healthcare resources on continuing chemotherapy at EOL. Thus, further investigation is warranted to explore the impact of regional healthcare resources at the residence level on continuation of chemotherapy at EOL for Taiwanese terminally ill cancer patients.

The strength of our study lies in its population-based sample, which encompassed all age and cancer groups of Taiwan. However, our study has several limitations. Continuation of chemotherapy was assessed using administrative data, which can be an imprecise approach because it depends on the quality of coding. Administrative data do not contain important clinical features that define the extent and aggressiveness of cancers, such as stage, tumor biology, and the intent of chemotherapy administered. The appropriateness of continuing chemotherapy in the last month of life may depend on an accurately defined population of patients dying from unexpected complications of curative chemotherapy for appropriate indications or dying of advanced cancer. Providing palliative chemotherapy at EOL for symptom relief may also be appropriate for some terminally ill cancer patients. This study did not address cancer patients’ preferences for using chemotherapy in advanced disease and other health outcomes, such as QOL, satisfaction with care, and healthcare costs to evaluate the appropriateness of continued chemotherapy at EOL.

Interpretation of our results may also have been limited by lack of detailed information on physicians’ characteristics (e.g. demographics, post-graduate training, knowledge and attitudes toward palliative chemotherapy). Despite using rigorous statistical methods and including a broad range of potential predictors, observational studies can never exclude the possibility of residual confounders. Further investigation should prospectively focus on understanding what drives variations in physician-, hospital-, regional-, and international-level continuation of chemotherapy in the last month of life. Special efforts may be focused on international differences in physicians’ training in palliative care, communication skills, and attitudes toward disclosing prognosis and directly discussing a transition from curative to palliative care with terminally ill cancer patients.

In countries with cultures of “family consent for disclosure” and “family autonomy” such as in Taiwan, patient prognosis tends to be disclosed to families, and family members have legitimate superior decision-making authority even over fully competent patients. However, our previous study indicated that Taiwanese family caregivers adopted a significantly more aggressive attitude toward life-sustaining treatments for their terminally ill relatives than the patients’ own stated preferences [34]. Terminally ill cancer patients without adequate knowledge of their prognosis predisposed them to overestimate their survival probability, leading them to prefer more aggressive EOL care [35]. Whether patients in countries with such family-dominant cultures are more likely to continue chemotherapy closer to death is an interesting topic for future study.

In conclusion, the proportions of Taiwanese cancer patients who continued chemotherapy in the last month of life were at the low end of worldwide statistics and remained steady from 2001 to 2006. The relative risk for continuing chemotherapy in the last month of life was determined by patient demographics and disease characteristics, physician specialty, and healthcare resources at the primary hospital level. Maintaining a delicate balance between deciding to continue chemotherapy and starting palliative care remains a challenge since predicting an individual cancer patient's life expectancy is often inaccurate, and desperate patients and families may prefer to continue chemotherapy to avoid death at any cost. Patients’ and families’ over-estimating the effectiveness of chemotherapy may be minimized by integrating palliative care into routine oncologic care. Reaching this goal may require a palliative care hospital-consultation team to facilitate adequate prognosis disclosure and appropriate discussion of the transition from curative to palliative care, thus achieving EOL care that best meets cancer patients’ needs and preferences.

Acknowledgements

Bureau of Health Promotion, Department of Health, Taiwan, R.O.C. (DOH96-HP-1510) with partial support from National Health Research Institute (NHRI-EX100–9906PI) and National Science Council (NSC 99–2628-B-182-031-MY2). No financial or other conflict of interest was involved in this study. None of the funding sources had any role in designing and conducting the study: collecting, managing, analyzing, and interpreting the data; or preparing, reviewing, or approving the manuscript.

Declaration of interest: The authors report no conflicts of interest. The authors alone are responsible for the content and writing of the paper.