INTRODUCTION
The application of aggressive end-of-life (EoL) care for cancer patients has recently become more and more popular (Cooke et al., Reference Cooke, Feemster and Wiener2014; Ho et al., Reference Ho, Barbera and Saskin2011). Discussions between terminal cancer patients and their physicians about preferences for EoL care can lead to such care being less aggressive (Wright et al., Reference Wright, Zhang and Ray2008; Mack et al., Reference Mack, Weeks and Wright2010). The current guidelines recommend that physicians discuss EoL care planning with terminal cancer patients early on in the course of the disease, during periods of physical and mental stability, rather than when the patient is undergoing acute deterioration (Peppercorn et al., Reference Peppercorn, Smith and Helft2011; Smith et al., Reference Smith, Temin and Alesi2012). Patients who have discussed their preferences for EoL care with their physicians are more likely to choose less-aggressive palliative care (Wright et al., Reference Wright, Zhang and Ray2008; Weeks et al., Reference Weeks, Cook and O'Day1998), and such less-aggressive care has been associated with better quality-of-life (QoL) close to the time of death (Wright et al., Reference Wright, Zhang and Ray2008; Reference Wright, Keating and Balboni2010).
When EoL care planning is discussed with patients early on in the course of their disease, care plans can be modified over time during disease progression (Straton et al., Reference Straton, Wang and Meoni2004; Fried et al., Reference Fried, Byers and Gallo2006; Wittink et al., Reference Wittink, Morales and Meoni2008). Even after advanced directives have been signed, patients can change their minds and update those directives. EoL care planning may be modified for various reasons, which can determine aggressiveness of care and quality of death (Earle et al., Reference Earle, Landrum and Souza2008; Wright et al., Reference Wright, Zhang and Ray2008). Although much previous research has investigated the stability of long-term EoL care preference in both the general population and specific clinical settings (Mukamel et al., Reference Mukamel, Ladd and Temkin-Greener2013; Barrio-Cantalejo et al., Reference Barrio-Cantalejo, Simon-Lorda and Molina-Ruiz2013; Janssen et al., Reference Janssen, Spruit and Schols2012; Cotter et al., Reference Cotter, Simon and Quinn2009; McParland et al., Reference McParland, Likourezos and Chichin2003; Weissman et al., Reference Weissman, Haas and Fowler1999; Danis et al., Reference Danis, Garrett and Harris1994; Everhart & Pearlman, Reference Everhart and Pearlman1990), there are very little data regarding short-term fluctuations in EoL care planning in seriously ill patients (Rosenfeld et al., Reference Rosenfeld, Wenger and Phillips1996). Identifying the factors related to these changes in patient preference for EoL care in terms of QoL could help improve the decision-making process in EoL care.
We evaluated the stability of terminally ill cancer patients' preferences regarding EoL care over time. We hypothesized that serial changes in QoL over time could affect EoL care planning, thus representing changes in attitude toward intensive care unit (ICU) care. We investigated the stability of preferences regarding ICU care and identified the factors associated with changes in QoL preferences.
PATIENTS AND METHODS
Study Design and Recruitment
The Study to Understand Risks, Priority, and Issues at the End of Life (SURPRISE), a prospective cohort study, enrolled terminal cancer patients in the Republic of Korea. Patients were recruited from 11 university hospitals and the National Cancer Center during the period from July of 2005 to October of 2006. Patients were eligible to participate if they were 18 years or older, had been diagnosed as terminal by their physicians, were capable of communicating with an interviewer or filling out questionnaires, and had been deemed competent to understand the purpose of the study and provide informed consent. They were enrolled within days of being diagnosed and informed of being in a terminal state. We defined a terminal patient as one with progressive, advanced disease who was likely to die within months. Patients were not eligible to participate if they continued anticancer treatment after enrollment, their disease status was non-evaluable, had changed their treatment plan, or could not complete the questionnaire because their condition had deteriorated. All participants provided informed consent, and our institutional review boards approved the protocol. The details of the study design have been published (Yun et al., Reference Yun, Kwon and Lee2010a ; Reference Yun, Lee and Kim2011).
We identified 481 terminal cancer patients who completed baseline questionnaires in SURPRISE (Figure 1). Of these, 473 were followed throughout their EoL period. During the 2-month follow-up, 330 patients died, and 2 were deemed incompetent. The data from 141 patients were subsequently analyzed.
Fig. 1. Flowchart of patient recruitment.
Data Collection
SURPRISE collected demographic data and clinical information about the patients. Questionnaires were administered within days of initiation of the study by face-to-face interviews. The same questionnaires were administered two months later by mail and took about 20 minutes to complete. SURPRISE gathered demographic information (i.e., age, sex, education level, religion, marital status, employment details, the person paying for treatment, and monthly household income) to evaluate previous discussions with physicians and patients about intentions regarding ICU care near death and other EoL issues. Following a previous investigation (Wittink et al., Reference Wittink, Morales and Meoni2008), the response on ICU care was coded as the dichotomous variables reject (“No, I would not want”) or accept (“Yes” or undetermined).
To measure QoL, we employed the validated Korean version of the European Organization for Research and Treatment of Cancer instrument (EORTC QLQ–C30) (Aaronson et al., Reference Aaronson, Ahmedzai and Bergman1993; Yun et al., Reference Yun, Park and Lee2004). The EORTC QLQ–C30 is a brief, self-reported, cancer-specific measure of QoL and is comprised of five, multiitem, functional scales that evaluate physical, role, emotional, cognitive, and social function, and one global health status/QoL scale. The three symptom scales measure fatigue, pain, and nausea/vomiting, and the six single items assess other symptoms (i.e., dyspnea, insomnia, appetite loss, constipation, and diarrhea) and financial difficulties. According to the EORTC scoring manual, all scales are then linearly transformed into a numerical score (0 to 100), with 100 representing the best global health status or functional status, or the worst symptom status, as appropriate. We handled incomplete questionnaires according to the developer's recommendations. We defined a meaningful difference in health-related QoL as a 10-point difference in mean score (Osoba et al., Reference Osoba, Rodrigues and Myles1998).
Each variable was categorized as follows: monthly income was categorized into “ < 2,000 US dollars (USD) and “ ≥ 2,000 USD”; education level into “high school or lower” and “college or beyond”; marital status into “married” and “unmarried,” which included “single” and “divorced/separated/widowed”; religion into “no religion” and “professing a religion,” including non-Catholic Christians, Buddhist, Catholics, and others; and the person paying for treatment into “patient” and “others,” including spouse, parents, offspring, relatives, and others. Patients also provided clinical information regarding their awareness of the terminal diagnosis and Eastern Cooperative Oncology Group (ECOG) performance status. ECOG performance status is an observer-rated scale of a patient's physical ability that utilizes numbers ranging from 0 (able to carry out all normal activities) to 4 (completely disabled) (Oken et al., Reference Oken, Creech and Tormey1982). We divided patients into two groups: those with scores of 0–2 and those scoring 3–4.
Statistical Analysis
Overall preference stability was assessed by kappa statistics. The primary outcome was the direction of the changed preference for ICU care near death. We categorized subjects into four groups according to changed preference: accept–accept group, reject–reject group, change to accept group, and change to reject group. Assuming that patients who accepted ICU care at baseline would have different reasons to change their preference at follow-up compared with patients who rejected ICU care at baseline, we compared patients in the accept–accept and accept–reject groups with those in the reject–reject with reject–accept groups (Rosenfeld et al., Reference Rosenfeld, Wenger and Phillips1996).
We employed a logistic regression model in order to assess the impact of patients' QoL changes or demographic/clinical characteristics on a shift of preference for ICU care near death. In addition, each independent factor that was statistically significant (p ≤ 0.10) in the univariate analysis was entered into the final multivariate logistic regression model to calculate adjusted odds ratios (aORs). Using a forward stepwise elimination procedure, we obtained a best-fit multivariate logistic regression model. In multivariate logistic analyses, we considered p values less than 0.05 generated in two-tailed tests to indicate statistical significance. All statistical tests were performed using SAS software (v. 9.2) (SAS Institute, Cary, North Carolina).
RESULTS
Patient Characteristics
Table 1 summarizes the demographic and clinical characteristics of participants. The mean age was 56.2 years (range 20–79 years), and nearly 50% were male. More than half of our subjects were aware of their terminal status (58.7%), and a large proportion of patients had not discussed ICU care near death with their physicians (79.7%).
Table 1. Characteristics of patients
ECOG = Eastern Cooperative Oncology Group.
USD = United States dollars.
ICU = intensive care unit.
Changes of Preference for ICU Care
Figure 2 depicts preference stability for ICU care during the 2 months prior to death (κ = 0.33, p < 0.001). In brief, the preference for ICU care during this time period did not change for 94 patients (66.7%; 48 for keep–accept and 46 for keep–reject), shifted to accept in 24 patients (17.0%), and shifted to reject in 23 patients (16.3%). At baseline, answers were 32 for “yes,” 70 for “no,” and 39 for “undetermined.” Among the 32 patients who answered “yes,” only 10 (31.3%) replied “yes” after the two-month follow-up (22 switched to “no” or “undetermined”). For those who answered “no” initially, the second answer was “yes” in 10%, “no” in 65.7%, and “undetermined” in 24.3%. Regarding subjects who answered “undetermined” at baseline, the response shifted to “yes” in 23.1% and to “no” in 30.8% (data not shown).
Fig. 2. Stability of ICU preference 2 months before death (N = 141). Degree of agreement was 66.7% (κ = 0.33, p < 0.001) by kappa statistics. ICU = intensive care unit.
Impact of QoL Changes on the Direction of Altered Preference
The data presented in Table 2 show the relationships between QoL changes and change in preference for ICU care near death. Patients who had maintained physical function changed their preference toward accept (OR = 0.34, p = 0.086), whereas those whose global health had deteriorated (OR = 2.37, p = 0.095) or those with non-aggravated fatigue (OR = 0.31, p = 0.031) altered their preference toward reject.
Table 2. Association between QoL changes* and preferences for ICU care
QoL = quality of life.
ICU = intensive care unit.
OR=odds ratio.
CI = confidence interval.
* More than 10 points.
Factors Related to the Altered Preference for ICU Care
Table 3 lists the demographic or clinical factors significantly associated with a changed preference for ICU care near death. Married patients were more likely to change their preference (OR for accept 0.10; OR for reject 0.22). Patients who were unaware of their terminal diagnosis tended to change their preference toward accept (OR = 2.49, p = 0.083), whereas those with poor performance were more likely to change their preference toward reject (OR = 2.45, p = 0.100).
Table 3. Characteristics related to a change in preference for ICU near death
ICU: intensive care unit.
USD = United States dollars.
ECOG: Eastern Cooperative Oncology Group.
OR=odds ratio.
CI = confidence interval.
Multivariate Analyses of Factors Related to Changes in Preference Regarding ICU Care
Table 4 shows the results of a multivariate analysis of factors associated with preferences regarding ICU care. Married patients were more likely to change their preference regarding ICU care (aOR toward accept 12.35, p = 0.021; aOR toward reject 7.87, p = 0.020). The likelihood of switching to reject was significantly higher in patients with poor performance (aOR = 5.32, p = 0.018), worsened QoL (aOR = 8.34, p = 0.007), or non-aggravated fatigue (aOR = 8.36, p = 0.006). The maintenance of physical function was a significant predictor of a change in preference toward accept (aOR = 5.05, p = 0.023).
Table 4. Predictors of a change in preference for ICU near death
ICU = intensive care unit.
QoL = quality of life.
aOR=adjusted odds ratio.
CI = confidence interval.
DISCUSSION
In this multicenter prospective study, we found that patient attitudes regarding ICU care at the end of life changed over time, and that deteriorating function was associated with this change. When patients were able to maintain physical function during their final two months, they tended to change their preference toward ICU care. However, when patients had worsening global health or stable symptoms such as fatigue, they tended to change their preference against ICU care. Interestingly, married patients were more likely to change their preference in either direction.
To the best of our knowledge, there is very little research on changes in preference for EoL care in seriously ill patients (Janssen et al., Reference Janssen, Spruit and Schols2012; Rosenfeld et al., Reference Rosenfeld, Wenger and Phillips1996; Fried et al., Reference Fried, Van Ness and Byers2007). The treatment preferences of patients who are confronting death should be given priority over those of healthy individuals imagining some future illness (Ditto et al., Reference Ditto, Jacobson and Smucker2006). A recent study identifying predictors of preference changes regarding life-sustaining treatments reported that decreased health status and mobility were associated with a change in preference (Janssen et al., Reference Janssen, Spruit and Schols2012), whereas we found that a deterioration in global QoL and physical function were related to a preference to reject ICU care. This discrepancy could be explained by the course of the disease and/or duration of follow-up, as the cited study investigated one-year stability of preferences in patients with advanced chronic organ failure. Since the functional decline associated with chronic organ failure is gradual compared to patients with terminal cancer, the reported preferences may be influenced by the individual's affective state rather than situational factors.
In our study, preferences toward ICU care at the EoL were significantly influenced by aggravated symptoms. Patients with stable symptoms tended to reject ICU care at the end of life. Patients who suffered from symptoms appeared to regard ICU care as another type of intensive care for management of their symptoms. However, actual ICU care at the EoL has been found to not be associated with prolonged survival (Yun et al., Reference Yun, Lee and Kim2011), and terminal cancer patients have been found to not benefit from ICU care (Kim et al., Reference Kim, Kim and Cho2014). Hence, sufficient management of symptoms is important not only for quality of life but also for appropriate EoL care planning.
Since patients who had earlier discussions about EoL care received less-aggressive treatment before death (see Mack et al., Reference Mack, Cronin and Keating2012), early EoL care planning should be undertaken in terminal cancer patients. The same study found that use of aggressive care was much less frequent when EoL discussions took place before the last 30 days of life, and that the odds of hospice utilization were nearly twice as high. We found that patient attitudes toward ICU care at the end life changed over time over the course of the disease, and that a preference for ICU care near the end of life changed in a third of terminally ill cancer patients. These results suggest a need for continuous discussions regarding EoL care planning as patient preferences for ICU care continue to change.
Generally, married patients have been found to be more likely to make predictable treatment decisions (Nielsen et al., Reference Nielsen, Mehlsen and Jensen2013; Erci & Ozdemir, Reference Erci and Ozdemir2009), which might be limited to when their conditions are stable (Janssen et al., Reference Janssen, Spruit and Schols2012). In our cohort of terminal cancer patients at the end of life, married patients were very vulnerable in the face of decisions regarding ICU care. Especially in Korea, an individual's health decisions are frequently made within a strong family context, where interdependence among family members is a prime value in decision making (Kwak & Salmon, Reference Kwak and Salmon2007; Mo et al., Reference Mo, Shin and Woo2012). Stronger interrelationships within a patient's family tend to have a great deal of influence on medical decisions at the end of life when the patient's condition is highly unstable.
Our study has several limitations. First, our participants were enrolled from selected university hospitals and might not represent the general population of terminal cancer patients. Nevertheless, our large multicenter-based setting and high participation rate should have minimized selection bias. Second, among our 481 participants, serial data for only 141 participants were available, for various reasons, introducing a potential selection bias. Third, we serially evaluated attitudes toward ICU care as part of EoL care planning. However, the intention to employ the ICU does not reflect all aspects of EoL care planning. Finally, other potential confounding factors were not considered (i.e., the influence of caregivers or the psychological aspects of individual patients) (Rosenfeld et al., Reference Rosenfeld, Wenger and Phillips1996; Yun et al., Reference Yun, Lee and Chang2010b ).
In conclusion, patient attitudes regarding treatment in an intensive care unit at the end of life are not stable over time, and changes in symptoms or function are associated with a tendency to changing attitudes. To improve medical decision making for EoL care, we believe that it is necessary to continue and broaden discussions about EoL care planning over time and thereby hopefully provide adequate care for cancer patients to preserve quality of life.
DISCLOSURES AND ACKNOWLEDGMENTS
This work was supported by National Cancer Center Grants 0410160 and 1310250-1. The authors state that they have no potential conflicts of interest to report.
