Introduction

In 2004, more than 65,000 people died of cancer in Korea, accounting for 26.3% of all deaths and making cancer one of the leading causes of death [1]. Cancer accounts for a significant proportion of healthcare spending in Korea. The total economic cost of cancer, including all cost components, was estimated to be $9.4 billion (1.72% of the gross domestic product) in Korea in 2002 [2]. Both cancer patients and their families suffer emotional trauma, a reduced quality of life, risk of premature death, and financial loss. Therefore, cancer is a major public health issue and creates a significant burden for patients, their families, and Korean society as a whole.

In contrast to the western world, few studies have examined cost outcomes of palliative care in Korea, where palliative and hospice care is not yet established. Hospitals are an appropriate setting for palliative care services because approximately 50% of adult deaths occur in hospitals, and virtually, all people with serious illness spend at least some time in a hospital during the course of their disease [3]. In Korea, we have seen a significant increase in the proportion of deaths in hospitals over the past 10 years [4]. The increase is disproportionately large for cancer (76.7%) compared with other diseases (54.7%) [4].

The purpose of this study was to estimate inpatient and outpatient medical cost per month for patients near the end-of-life and evaluate the cost-saving effect of palliative care. Medical expenses of Korean patients within 6 months of death were compared for those admitted to the palliative care unit (PCU) and those not admitted to the palliative care unit (non-PCU), thus allowing determination of demographic or clinical clues related to cost savings.

Patients and methods

We evaluated data from a retrospective cohort composed of 709 patients who died as a result of cancer at Seoul National University Hospital (SNUH) from January 1 to December 31, 2007. Of those subjects, 53 were excluded from the study due to age (32 children) and/or insufficient medical information (21). Data from the remaining 656 patients were included in the analysis. Of those patients, 126 died in the PCU, 461 died in a non-PCU, and 69 died in the emergency room (ER) before PCU or non-PCU admission.

The admission criteria to a PCU are as follows: (1) patients who need palliative care, (2) patients who are in a stage in which anticancer therapy is no longer beneficial due to disease progression or poor performance. The proportion of terminal cancer patients in the PCU is 80%–90%.

Patients are supposed to be admitted to the neurology or psychiatry ward prior to hospitalization in cases of serious psychosis, delirium, dementia, or depression. Although cognitive function tests are not performed routinely, neuropsychiatric consultations were conducted for patients who showed cognitive function deterioration during their stay in the PCU ward, and they were included in this study.

Information related to the patients' medical costs for 6 months prior to death was collected. These costs included hospital charges and outpatient charges. The data sources for this study were health insurance reviews, assessment services, and medical records from SNUH. Demographic characteristics evaluated included age, sex, site of death, diagnosis, education level, residence, and marital status. To determine the cause of increased medical expenses, we retrospectively reviewed medical records, for clinical information associated with medical cost, such as do-not-resuscitate (DNR) permission, intensive care unit (ICU) admission, ER visits, ventilator use, chemotherapy, surgery, hemodialysis, total parenteral nutrition (TPN) within 2 months of death, and admission days within 6 months of death. In this study, the number of personnel working at the PCU was the same as those in the general wards. In the case of other health technologies, medical costs for patients are based on a fee-for-service system (including personnel expenditures). More volunteers are contributing in the PCU for non-medical activities, but this cost was not counted.

The total economic cost was calculated by combining the separate estimates of cancer costs. All costs estimated were represented using an exchange rate of 950 Korean won to 1 US dollar, which was the annual exchange rate in 2007.

A t test or chi-square test was used to compare the characteristics between PCU and non-PCU patients within 6 months of death. Repeated measures ANOVA was used to examine whether changes in subscales of medical expenses were statistically significant.

Plausible variables based on previous reports and our hypothesis were entered into a regression model for predicting changes in medical costs. We considered P < 0.05 as statistically significant and reported two-sided P values.

Results

Clinical characteristics

The clinical characteristics of included patients are shown in Table 1. Men (n = 407) and women (n = 180) with a median age of 61 years (mean 58.7; range 17–96 years) were evaluated. Patients who died in the emergency room (n = 69) were excluded.

Table 1 Demographic and clinical characteristics of patients
Full size table

The PCU group included 68 men and 58 women with a median age of 59 (mean 56; range 22–79 years), and the non-PCU group included 339 men and 122 women with a median age of 61 (mean 59.5; range 17–96). Female and younger patients were found to use the PCU more frequently (P < 0.001, P = 0.002). Three hundred patients died in the general ward (51.1%); 161 died in the ICU (27.4%); and 126 died in the PCU (21.4%).

Diagnosis were hepatoma (25.9%), lung cancer (15.2%), stomach cancer (11.2%), lymphoma (7.0%), and leukemia (6.5%). In the PCU group, stomach cancer was most common (26.2%), followed by lung cancer (23.8%), and breast cancer (10.3%). In the non-PCU group, there were 151 cases of hepatoma (32.8%), 59 lung cancer (12.8%), and 37 leukemia (8.0%). There was no statistical difference between the two groups in terms of education level, marital status, or residential area. There was also no difference in the mean hospital stay within 6 months of death between PCU (45.90 ± 34.94) and non-PCU patients (47.05 ± 39.5).

The PCU group had higher rates of DNR, ER visits, and palliative chemotherapy treatment (P < 0.001, P = 0.045, P = 0.002), while the non-PCU group had higher frequencies of using the ICU, ventilators, and hemodialysis (P < 0.001, P < 0.001, P < 0.001). There were no statistical differences between the two groups in terms of TPN and surgery.

Medical cost according to PCU usage

The average medical cost per patient within 6 months of death was $27,863, composed of $24,799 in hospital charges, and $3,063 in outpatient and emergency room charges. The total medical cost within 6 months of death for the PCU group was significantly lower than the non-PCU group ($21,591 vs. $29,577, P < 0.001). This difference was due to higher hospital charges for the non-PCU group in spite of their lower outpatient and emergency room costs. The non-PCU group paid significantly higher admission fees which increased their medical expenses per day.

Medical costs per month during the 6 months prior to death dramatically increased as patients reach the terminal stage (Fig. 1). There was a significant difference in the monthly medical cost during the last 2 months between PCU and non-PCU patients. Medical cost in the last month of life for non-PCU patients was $12,428, which accounts for 44.6% of the total medical costs within 6 months of death. The PCU group spent 32.8% less in the final month and 33.0% less in the month prior than the non-PCU group (P = 0.004, P < 0.001). (Table 2)

Fig. 1
figure 1

a Distribution of medical expenses in the last 6 months of life according to the months before death. b Comparison of distribution of medical expenses in the last 6 months of life according to palliative care

Full size image
Table 2 Differences in total medical expenses and monthly expenses between palliative care and non-palliative care
Full size table

Medical cost according to clinical characteristics

Patient gender had no effect on medical costs. Patients <30 years of age spent more on medical expenses than patients ≥70 years of age and 30–69 years of age (P < 0.001). As shown in Tables 1 and 3, the main population difference was in the age groups 30–69 and ≥70, but no difference was observed for medical expenses in these groups. Young patients (<30 years) spent more on healthcare costs, but the number of patients was similar in the PCU and non-PCU groups. Medical costs, according to place of death, were higher in the ICU ($33,063) followed by the general ward ($27,706), and the PCU ($21,591) (P < 0.001). Medical costs of those with hematologic diseases, high school–graduate level education, and unmarried were significantly higher (P < 0.001, P < 0.001, P < 0.001).

Table 3 Medical expenses according to demographic and clinical characteristics
Full size table

For medical cost within 6 months of death, the longer the length of stay, the greater the expense (P < 0.001). Patients who were admitted to the ICU within 2 months of death, and had used a ventilator, received chemotherapy, TPN, surgery, or hemodialysis paid more medical expenses. However, the DNR decision had no impact on medical expenses. Patients who visited the ER within 2 months before death had lower medical expenses (P < 0.001) (Table 3).

Factors influencing medical costs

We conducted regression analysis to evaluate the factors influencing total medical cost within 6 months of a patient's death. The definition of variables and measurements are shown in Table 4. Four regression models were constructed with demographic variables, site of death, and clinical factors: model 1 consisted of demographic variables only, model 2 included demographic variables and site of death, model 3 considered both site of death and clinical characteristics, and model 4 included all relevant factors (Table 5).

Table 4 Definition of variables
Full size table
Table 5 Regression results of factors affecting medical expenses of patients during the 6 months prior to death
Full size table

The regression models were as follows:

$$ Model 1:Y = {\beta_1} + \sum_{{{\rm{i}} = 2}}^7{\beta_{\rm{i}}}{X_{\rm{i}}} + {u_i} $$
$$ {\hbox{Model 2}}:Y = {\beta_1} + \sum_{{{\rm{i}} = 2}}^7{\beta_{\rm{i}}}{X_{\rm{i}}} + \sum_{{{\rm{i}} = 8}}^{{10}}{\beta_{\rm{i}}}{X_{\rm{i}}} + {u_i} $$
$$ {\hbox{Model 3}}:Y = {\beta_1} + \sum_{{{\rm{i = 8}}}}^{{10}}{\beta_{\rm{i}}}{X_{\rm{i}}} + \sum_{{{\rm{i}} = 11}}^{{20}}{\beta_{\rm{i}}}{X_{\rm{i}}} + {u_i} $$
$$ {\hbox{Model }}4:Y = {\beta_1} + \sum_{{{\rm{i}} = 2}}^7{\beta_{\rm{i}}}{X_{\rm{i}}} + \sum_{{{\rm{i}} = 8}}^{{10}}{\beta_{\rm{i}}}{X_{\rm{i}}} + \sum_{{{\rm{i}} = 11}}^{{20}}{\beta_{\rm{i}}}{X_{\rm{i}}} + {u_i} $$
Y :

Total medical cost

X i :

Explanatory variables

β i :

Parameters

u i :

Disturbance

Regression results indicated that demographic variables had little relationship to medical expenses in model 1. The site of death variable increased coefficients of determination from 0.040 in model 1 to 0.115 in model 2. After incorporating the site of death variables and clinical factors in models 3 and 4, however, coefficients of determination sharply increased to 0.574 and 0.576, respectively. Medical cost within 6 months of death were higher in patients who had stayed longer in the hospital, underwent mechanical ventilation within 2 months, and received chemotherapy, TPN, or hemodialysis (P < 0.05). Patients who used the PCU, had solid tumors, and had less than a high school education, paid lower medical expenses (P < 0.05) (Table 5).

Discussion

Palliative care is a multidisciplinary approach which aims to relieve suffering and improve the quality of life for patients with advanced life-limiting illnesses and their families. Palliative care is intended to enhance patient care and help alleviate the financial burden of patients with serious life-limiting illnesses [5]. However, switching terminally ill patients from the ICU to the PCU sometimes initiates a conflict between medical and cultural beliefs. Discrepancies can exist among patients, family members, and physicians regarding appropriate life-sustaining treatment [6, 7]. There are complex problems to consider when moving terminally ill patients to comfort care situations. Family members in Korea do not usually want patients to be informed of the exact status of their disease and sometimes do not even want patients to be informed at all that they have a malignant disease [6, 8]. These cultural obstacles and taboos in Korea and other Asian countries contribute to the lack of shared decision making between physicians and end-of-life cancer patients. In addition, treatment limitations are often not discussed with patients [9]. Family caregivers under Confucian's culture often have a significantly more aggressive attitude toward end-of-life care than the patient because of the obligations they feel towards their family [10, 11]. Moreover, institutionalization of palliative care has not yet been established, and as such, there is no separate PCU within a general hospital in Korea. The availability of end-of-life care facilities in rural districts of Korea is currently limited. Only 6.3% of total deaths from cancers used hospice palliative care facilities during 2008, and health insurance does not cover hospice care service. In addition, most patients want to stay in major hospitals until death. One of the purposes of this study was to develop a transfer hub system from active anti-cancer treatment in major hospitals to local hospice units.

A large increase in the proportion of deaths occurring in hospitals leads to inappropriate use of hospital resources and produces unnecessary medical costs. In this study, a significant proportion of patients received aggressive treatment up to the end-of-life. The average total medical cost within 6 months of death was $27,863 which was mainly due to inpatient costs and is higher than previous reports whose costs considered insurer payments and co-payments [12]. There are a number of reasons for this increase in patient medical expenses: first, the total medical expenses in this study involved total costs not only those covered but also those not covered by insurance; second, SNUH is a tertiary referral hospital, in which patients tend to receive more aggressive care and chemotherapy [13]; and third, the number of patients receiving anti-cancer treatments in the final stages is increasing [14]. PCU patients received palliative chemotherapy significantly more than non-PCU in our study. One reason for this difference is that 32% of the non-PCU group had hepatocellular carcinoma, which rarely responds to palliative chemotherapy. The Korean FDA approved sorafenib for hepatocellular carcinoma in 2008. The patients for this study were enrolled from January 1 to December 31, 2007. In this study, biologics did not influence medical costs significantly. However, medical costs due to the use of biologics (monoclonal antibodies, etc.) in the end-of-life period are a big issue in this country.

Palliative care in hospitals is associated with significant reductions in per diem costs and total costs, and can generate substantial savings to the health system by “cost avoidance” [15]. In our study, the PCU was an independent factor involved in the reduction of medical costs shown to decrease total medical expenses by 27% within 6 months of death. Medical costs per month between the two groups evaluated increased in the last 2 months before death. Consistent with a previous report, the cost-reduction effect of palliative care is noticeable as the patient reaches the final stage [16]. Medical costs are also influenced by physicians' decisions regarding the patients' terminal stage. Other factors found to be associated with reduced medical expenses were solid cancers and lower educational levels. Factors increasing medical costs were length of hospital stay, ventilator use, chemotherapy, TPN, and dialysis. DNR permission did not impact medical expenses.

The limitations of our study are the following: (1) this study is retrospective in design, and other studies should be performed prospectively for validation; (2) it represents only a single center and is subject to selection bias as it reflects the medical cost data from only one institution. The results may not be generalizable. This study should be validated in a multi-center trial; (3) we did not include quality of life (QOL) data, therefore, our results do not show to what extent medical costs influence QOL. Although QoL data are not collected prospectively, most physicians evaluate QoL to determine anti-cancer treatments. However, an ethical dilemma exists for withholding or withdrawing anti-cancer and life-sustaining treatments. Most patients and family members want to do everything possible until the end, although many times, this is futile. Both medical education and social campaigns are necessary to improve the situation; (4) medical care costs in our study do not represent the total economic costs of cancer but only the direct medical costs. Nevertheless, this study is the first to report on the role of the PCU in hospital medical costs and other cost-reducing factors related to the care of terminal cancer patients in Korea.

In analyzing patients' last 6 months of medical expenses, the cost-saving effect of palliative care was verified and some specific factors tied to cost increases during this time period were explored. If the parameters given in this study are controlled properly in the era of institutionalization of palliative care, it is expected that effective cost-savings could be realized.