Transitions in Prognostic Awareness Among Terminally Ill Cancer Patients in Their Last 6 Months of Life Examined by Multi‐State Markov Modeling

Accurate prognostic awareness is a cornerstone of preference‐based end‐of‐life care decision‐making for terminally ill cancer patients, promoting optimal quality of care at the end of life. This study explored the transition probabilities in distinct states of prognostic awareness in terminally ill cancer patients in the last 6 months of life.

Abstract

Background.

Developing accurate prognostic awareness, a cornerstone of preference‐based end‐of‐life (EOL) care decision‐making, is a dynamic process involving more prognostic‐awareness states than knowing or not knowing. Understanding the transition probabilities and time spent in each prognostic‐awareness state can help clinicians identify trigger points for facilitating transitions toward accurate prognostic awareness. We examined transition probabilities in distinct prognostic‐awareness states between consecutive time points in 247 cancer patients’ last 6 months and estimated the time spent in each state.

Methods.

Prognostic awareness was categorized into four states: (a) unknown and not wanting to know, state 1; (b) unknown but wanting to know, state 2; (c) inaccurate awareness, state 3; and (d) accurate awareness, state 4. Transitional probabilities were examined by multistate Markov modeling.

Results.

Initially, 59.5% of patients had accurate prognostic awareness, whereas the probabilities of being in states 1–3 were 8.1%, 17.4%, and 15.0%, respectively. Patients’ prognostic awareness generally remained unchanged (probabilities of remaining in the same state: 45.5%–92.9%). If prognostic awareness changed, it tended to shift toward higher prognostic‐awareness states (probabilities of shifting to state 4 were 23.2%–36.6% for patients initially in states 1–3, followed by probabilities of shifting to state 3 for those in states 1 and 2 [9.8%–10.1%]). Patients were estimated to spend 1.29, 0.42, 0.68, and 3.61 months in states 1–4, respectively, in their last 6 months.

Conclusion.

Terminally ill cancer patients’ prognostic awareness generally remained unchanged, with a tendency to become more aware of their prognosis. Health care professionals should facilitate patients’ transitions toward accurate prognostic awareness in a timely manner to promote preference‐based EOL decisions.

Implications for Practice.

Terminally ill Taiwanese cancer patients’ prognostic awareness generally remained stable, with a tendency toward developing higher states of awareness. Health care professionals should appropriately assess patients’ readiness for prognostic information and respect patients’ reluctance to confront their poor prognosis if they are not ready to know, but sensitively coach them to cultivate their accurate prognostic awareness, provide desired and understandable prognostic information for those who are ready to know, and give direct and honest prognostic information to clarify any misunderstandings for those with inaccurate awareness, thus ensuring that they develop accurate and realistic prognostic knowledge in time to make end‐of‐life care decisions.

Introduction

Accurate prognostic awareness is a cornerstone of preference‐based end‐of‐life (EOL) care decision‐making for terminally ill cancer patients [1], thus promoting optimal quality of care at EOL [2], [3], [4], [5]. Indeed, accurate prognostic awareness facilitates patients’ receiving fewer futile treatments at EOL [2], being prepared for death [3], and having better quality of life [4], [5]. However, Taiwanese physicians commonly do not disclose the prognosis to terminally ill cancer patients [6], which may lead to inaccurate prognostic awareness [6], [7].

Becoming accurately aware of one's poor prognosis is a dynamic process [8], [9], [10], [11], [12], [13], but in a recent 34‐study systematic review of advanced/terminally ill cancer patients’ prognostic awareness [14], the majority of studies (n = 22; 64.7%) used cross‐sectional surveys. Only a handful of longitudinal studies explored changes in patients’ prognostic awareness more than twice [9], [10], [11], [12], [13], [15], [16], with several methodological shortcomings. Few studies measured changes in cancer patients’ prognostic awareness until their death or last month of life [11], [12], [13]. Some explored individual‐level changes in prognostic awareness [9], [10], [12], [15], [17], [18], but none investigated transition probabilities in distinct states of prognostic awareness between consecutive time points.

Patients’ desire and need for prognostic information is a key to accurate prognostic awareness [19], [20]. However, desire for prognosis [21], [22] and prognostic awareness [9], [10], [11], [12], [13], [15], [16], [17], [18] are often treated as isolated events and explored separately. To understand the process that terminally ill cancer patients go through to reach accurate prognostic awareness, a useful conceptual framework is the transtheoretical model (TTM) [23], [24]. The TTM proposes that individuals undergo a health behavior change (e.g., acquire accurate prognostic awareness) by progressing from precontemplation (i.e., no desire or not being ready to know one's prognosis) and contemplation stages (e.g., seeking prognostic information [the desire to know one's prognosis]) to preparation, action, and maintenance stages (e.g., acquiring prognostic information and transitioning to accurate prognostic awareness) [23], [24].

Physicians tend to discuss prognosis with patients late in their illness trajectory [25], [26], [27], [28] resulting in late referrals to palliative care [29], despite early integration of palliative care into cancer care being recognized as the “gold standard” for patients with advanced/metastatic cancer [30]. To achieve this goal, understanding the time terminally ill cancer patients spend in each state of prognostic awareness is important to identify trigger points for facilitating transitions to higher prognostic‐awareness states, especially for patients who are ready to receive prognostic information to make EOL‐care decisions. Therefore, the purposes of this study were to longitudinally explore the transition probabilities in distinct states of prognostic awareness between consecutive time points in terminally ill cancer patients’ last 6 months of life and to estimate the time spent in each state.

Methods

Study Design and Sample

For this longitudinal study, a convenience sample of terminally ill cancer patients was recruited in 2010–2012 from the general medical inpatient units of a medical center in northwest Taiwan and followed through December 2013. Participants and data for this study were taken from a study on terminally ill cancer patients’ preferences for life‐sustaining treatments; sampling and data‐collection details have been published [31]. In brief, adult cancer patients were referred by their oncologist, who recognized them as terminally ill based on their disease continuing to progress and not responding to curative treatments. Participants were interviewed by experienced oncology nurses approximately every 2 weeks until they declined participation or died. To investigate the pattern of changes in prognostic awareness, we included patients with at least two assessments after enrollment. This study was approved by the ethics committee of the study site. All participants provided written informed consent.

Measures of Prognostic Awareness and Desire for Prognostic Information

Prognostic awareness was evaluated by asking patients if they knew their prognosis, and, if so, whether their disease (a) was curable; (b) might recur in the future, but their life was not currently in danger; and (c) could not be cured, and they would probably die in the near future [32]. Patients were recognized as accurately understanding their prognosis only if they chose option c. Patients were recognized as inaccurately understanding their prognosis if they chose options a or b. If patients indicated that they did not know their prognosis, they were then asked to rate their degree of desire for prognostic information (e.g., “Do you want your physician to tell you whether your disease can be cured or not?”) on a 5‐point Likert scale from 1 (very undesirable) to 5 (very desirable). Desire for prognostic information was dichotomized at the median score into want to know (≥3) and do not want to know (<3). Patients’ prognostic awareness was therefore categorized into four states: (a) do not know and do not want to know, state 1; (b) do not know but want to know, state 2; (c) inaccurate awareness, state 3; and (c) accurate awareness, state 4.

Analysis

Baseline characteristics and prognostic awareness of participants who died, withdrew, were still alive at the end of follow‐ups, were assessed only once, or were excluded from analysis (e.g., prognostic awareness was not assessed or data were collected more than 6 months before death) were compared by chi‐square tests and analysis of variance.

Transition probabilities in distinct states of prognostic awareness between consecutive time points and the time spent in each state were estimated by multistate Markov modeling (MSM) [33], [34]. MSM describes dynamic changes in patients’ prognostic awareness by simultaneously estimating initial prevalence for each state and the transition probabilities for patients in a specific state moving between states from time point t‐1 to time point t. Estimates of transition probabilities for patients’ prognostic awareness are derived from observed data, including assessment time, the state occupied at each assessment, and the observation scheme [33], [34].

Once transition probabilities were estimated, they were used to compute the time spent in each state if patients survived 6 months after enrollment (when they were initially recognized as terminally ill) [33], [34]. In our analysis, we considered all possible directional transitions in prognostic awareness between the 4 states of prognostic awareness. The validity of the model was tested by comparing graphs of observed and expected state prevalences at given times [33], [34]. All analyses were conducted using the MSM package in R (statistical software, R Project for Statistical Computing, Vienna, Austria, http://www.R-project.org/) [35], [36].

Results

Sample Characteristics

Of 433 eligible patients, 380 were enrolled (participation rate = 87.8%) (Fig. 1). Among enrolled patients, 100 were excluded from analyses because prognostic awareness was assessed only after they were enrolled (n = 37), their prognostic awareness was not assessed (n = 3), their data were collected more than 6 months before death (n = 8), their data were collected only once (n = 50), or they survived over 3 years after enrollment (not terminally ill; n = 2). At the end of follow‐ups, 21 patients withdrew and 12 were still alive.

Figure 1.

Participant flow chart.

The 247 patients who died and supplied sufficient prognostic awareness data comprised the final sample. The five patient groups (study sample, those who withdrew, those alive, those assessed only once, and those excluded from analysis [prognostic awareness not assessed, data collected more than 6 months before death, and survived more than 3 years after enrollment]) did not differ significantly in baseline sociodemographics (data not shown), except for disease burden and prognostic awareness. Patients excluded from analysis and those whose prognostic awareness was assessed only once suffered lower and higher disease burdens (symptom distress and functional dependence) than the sample, respectively. Patients who were still alive were less accurately aware of their prognosis than those who died during the enrollment period (data not shown).

Participants’ sociodemographic and disease characteristics are in Table 1. After enrollment, participants survived on average 164.5 days (SD = 143.2; median = 109; range = 17–667) and completed on average 5.9 follow‐up assessments (SD = 2.8; median = 5; range = 2–12; 87.9% > 2 assessments). The following results are based on 1,446 assessments, separated averagely by 18.1 days (SD = 7.0; median = 15; range = 4–82). The last assessment was made on average 25 days (SD = 23.6; median = 18; range = 1–137) before death.

Table 1. Participants’ demographic and clinical characteristics.

Transitions in Prognostic Awareness in Cancer Patients’ Last 6 Months of Life

At baseline, the most prevalent state of prognostic awareness was accurate awareness (state 4, 59.5%), followed by not knowing but wanting to know (state 2, 17.4%) and inaccurate awareness (state 3, 15.0%). Less than one‐tenth of participants neither knew nor wanted to know their prognosis (state 1, 8.1%).

The transition probabilities for patients in a specific state moving between states from time point t‐1 to time point t are presented in Table 2. Terminally ill cancer patients’ prognostic awareness generally remained unchanged (probabilities of remaining in the same state were 45.5%–92.9%), indicating stable prognostic awareness, especially for those in state 4. If prognostic awareness changed, it tended to shift toward higher (more aware) states of prognostic awareness (for those in states 1–3, probabilities of shifting to state 4 were 23.2%–36.6%, followed by 9.8%–10.1% probabilities for those in states 1 and 2 shifting to state 3). Based on each state's initial probability (“initial size”) and transition probabilities, we estimated the probabilities for each state in patients’ last 6 months of life since enrollment (recognition of terminally ill status; Fig. 2). As time passed and death approached, the proportion of patients with accurate prognostic awareness (state 4) increased prominently, whereas the prevalences of the other three states showed decreasing trends, especially for state 2. The estimated final prevalences of prognostic awareness right before patient death in states 1, 2, 3, and 4 were 3.6%, 5.0%, 9.6%, and 81.8%, respectively, if patients survived 6 months after enrollment.

Table 2. Transition probabilities of prognostic awareness from time [t‐1] to time [t] among cancer patients in their last 6 months.

Bold indicates the highest transition probability between time points.

Figure 2.

Estimated probabilities for each prognostic‐awareness state at different times after enrollment in patients’ last 6 months of life.

The estimated times that terminally ill cancer patients spent in prognostic‐awareness states 1, 2, 3, and 4 were 1.29, 0.42, 0.68, and 3.61 months, respectively, in their last 6 months of life. The validity of our model is supported by graphical comparison of the observed and expected prevalences of each state at given times (Fig. 3).

Figure 3.

Model fit between observed and expected prevalences over patients’ last 6 months of life: comparison for each state.

Discussion

The majority of our patients (59.5%) were accurately aware of their prognosis at baseline, with a prevalence higher than that (49.1%) synthesized in a previous meta‐analysis of advanced/terminally ill cancer patients [14]. Our higher observed initial prevalence of accurate prognostic awareness may be primarily due to our participants being diagnosed with terminal illness rather than the mix of meta‐analysis study subjects with advanced and terminal illness [14]. Terminally ill cancer patients are more likely than advanced cancer patients to accurately understand their prognosis because as death approaches, their physical condition deteriorates [37] and their physicians tend to disclose prognosis to them [25], [26], [28].

Our findings indicate that terminally ill cancer patients’ prognostic awareness generally remained unchanged (probabilities of remaining in the original state were 45.5%–92.9%), especially accurate prognostic awareness, consistent with reports [12], [17] that prognostic awareness remained accurate over time for the majority of terminally ill cancer patients. Prognostic awareness was mostly stable for our participants in state 4 (accurate prognostic awareness). According to the TTM, these patients are in the action and maintenance stages of prognostic acceptance [23], [24]. As death approaches, patients’ physical condition continues to deteriorate [20], [37] and, over time, they cognitively integrate prognostic information from discussions with their physicians [38], [39], thus gradually coming to realize their prognosis. Therefore, they are more likely to consolidate their prognostic awareness rather than to swing or regress back to earlier (less aware) stages.

We estimated that our participants spent 3.61 months in state 4 if they survived 6 months after enrollment when their terminal status was first recognized. When patients maintain accurate prognostic awareness for 3.61 months before death, they may have not only sufficient time to consider and reflect on their values and EOL‐care preferences, but also more opportunities to discuss those preferences with their physicians [40], [41], thus promoting preference‐based EOL‐care decision‐making [1], [2]. However, many of our patients (61.7%) with accurate prognostic awareness spent less than 3.61 months in state 4 before death (data not shown), possibly related to physicians’ late prognostic disclosure. Indeed, physicians often discuss prognosis/EOL care with patients late in their illness [25], [26], [27], [28], [40], [41]. For example, if physicians discussed prognosis with their patients, they reported often first disclosing prognostic information to patients in the last month of their life [25]. Physicians also commonly wait to discuss prognosis with terminally ill patients until they or their family members bring up the issue [26], [27], [28], or when curative treatments are no longer available [26], [28]. Therefore, to facilitate patients’ accurate prognostic awareness, U.S. national guidelines [42], [43], [44] recommend that physicians discuss prognostic information closer to patients’ diagnosis of advanced cancer, when they are cognitively competent to make informed and value‐consistent decisions regarding their EOL care.

We found that if prognostic awareness changed, patients in state 2 (do not know but want to know) had the highest probability (54.5%) of shifting toward other states. Of note, patients in state 2 had the greatest probability of shifting to state 4 (36.6%), followed by shifting to state 3 (9.8%). In addition, patients in state 2 spent the shortest time (0.42 month) in that state among the four states of prognostic awareness. According to TTM, patients in the contemplation stage (e.g., state 2 of our study) earnestly prepare for change by consciousness‐raising to gather related information [23], [24]. Terminally ill cancer patients in state 2 who want prognostic information may become receptive to more detailed information about their prognosis, thereby being more likely to accurately know their prognosis [19], [20]. In addition, physician disclosure of prognosis, a major facilitator of accurate prognostic awareness [39], may be initiated in response to patients’ desire and requests for prognostic information [7], [27]. Therefore, when health professionals perceive patient cues that they are willing and ready to know about their prognosis, professionals should cultivate this desire using an evidence‐based guide [42], [43], [44], [46] to provide desired and understandable prognostic information.

Patients in state 3 (inaccurate awareness) had a 33.9% probability of shifting to state 4 in the subsequent assessment, consistent with evidence that some patients (15%–28%) come to accurately understand their disease over time [12], [17], [18]. Patients in state 3 are considered by the TTM to be in the preparation stage, i.e., they are ready to attempt behavior change in the near future because they have already done so [24]. Terminally ill cancer patients in state 3 may be ready and prepared for prognostic information (supported by their spending the second shortest time in this state). However, patients may guess their prognosis from sources other than physician disclosure [7] because Taiwanese physicians commonly do not disclose prognostic information to patients [6]; such guessing leads to inaccurate prognostic awareness [7]. On the other hand, patients have a high chance of shifting to accurate prognostic awareness if health care professionals respond appropriately to their informational needs. Professionals should sensitively explore terminally ill cancer patients’ preparedness for prognostic information, directly and honestly communicate with them, and clarify any misunderstandings of prognostic information to avoid patient overestimation of prognosis [1] and unrealistic expectations of continuing anticancer treatments [47].

The most worrisome situation is for terminally ill cancer patients in state 1 (neither knowing nor wanting to know their prognosis) because they had the lowest probabilities of shifting to state 4 (23.2%) or state 3 (10.1%) and spent the second longest time (1.29 months) in state 1 than in other states. According to the TTM, patients in state 1 are in the precontemplation stage, in which individuals have been shown to be less likely to change their behaviors because they are often unaware of the benefits of change [23], [24]. Therefore, professionals should consider patients’ state of preparedness and provide appropriate interventions [46].

For example, health professionals have been suggested to defer giving prognostic information to patients reluctant to discuss prognosis and in stable condition; these patients should be repeatedly assessed for their understanding of the likely illness trajectory [46]. In contrast, for patients reluctant to discuss prognosis and physically deteriorating, professionals should raise patients’ prognostic awareness by encouraging them to consider their current condition and “naming the dilemma,” i.e., empathizing with their difficulty in thinking about possibly dying, but worrying that not discussing prognosis will lead to poor treatment decision‐making [46]. More sensitive innovative interventions are needed to address the prognostic information needs of terminally ill cancer patients in state 1, thus facilitating their transition to more accurate states of prognostic awareness for making appropriate EOL‐care decisions.

Our results contribute substantially to knowledge of terminally ill cancer patients’ accurate prognostic awareness between diagnosis of terminal status and death by using MSM. This modern analytic approach, which allowed us to make full use of longitudinal data at the individual level as death approached, provides an in‐depth understanding of the transition probabilities in distinct states of prognostic awareness between consecutive time points and the time spent in each distinct state.

Nonetheless, this study had several limitations. Participants were recruited by convenience sampling from a single medical center in Taiwan, limiting representativeness of the findings to national and international target populations. A noticeable proportion of patients withdrew or were excluded from analysis, limiting generalization of our findings to those patients. Our measure of prognostic awareness was adopted from published studies [48], [49], [50] and modified based on Taiwanese physicians’ cultural practice of disclosing prognosis, limiting its generalizability in other cultures and health care practices. However, this measure reflects common conceptualizations and measures of prognostic awareness used in a recent 34‐study systematic review [14]. Replicating our study method and validating our findings are warranted by using other measures of prognostic awareness, such as that developed to measure terminal illness acknowledgement [38]. We investigated whether prognostic awareness and desire for prognostic information were from patients’ perspective only, but did not evaluate the timing, depth, or quality of actual patient‐physician communication of prognostic information. Furthermore, we investigated patients’ states of prognostic awareness without taking family caregivers’ perspectives into consideration. However, in Asian countries such as Japan, Korea, and China/Taiwan, prognosis is commonly disclosed to the family instead of patients (“family consent for disclosure”) [51] based on the cultural value of filial piety, the belief that patients will be harmed by knowing their poor prognosis, and the relative power of family [52], [53]. Indeed, evidence shows that terminally ill Asian cancer patients and their family caregivers differ substantially in their knowledge and experiences of being informed about prognosis, with patients understanding their prognosis considerably less accurately than their family caregivers [20], [54], [55], [56]. Roles played by family caregivers in facilitating or impeding terminally ill cancer patients’ transitions toward higher states of prognostic awareness warrant further investigations to guide effective interventions to facilitate accurate prognostic awareness for Asian patients at EOL. Finally, we did not incorporate covariates into MSM to understand factors that may change state transition rates.

Conclusion

Terminally ill Taiwanese cancer patients’ prognostic awareness generally remained stable, but when prognostic awareness changed, it tended to shift toward higher states of awareness. Since terminally ill cancer patients need time to develop an accurate awareness of their prognosis, clinicians should begin early in the disease trajectory to cultivate patients’ accurate prognostic awareness by specific interventions [e.g., [46]] tailored to their readiness for prognostic information in each distinct state. Such timely and sensitive interventions would ensure that patients make full use of prognostic information in making informed and value‐consistent EOL‐care decisions. Health professionals should appropriately assess terminally ill cancer patients’ readiness for prognostic information and respect patients’ reluctance to confront their poor prognosis if they are not ready to know, but sensitively coach them to slowly cultivate their accurate prognostic awareness, provide desired and understandable prognostic information for those who are ready to know, and give direct and honest prognostic information to clarify any misunderstandings of prognosis for those with inaccurate awareness, thus ensuring that their knowledge is sufficiently accurate and realistic to make EOL‐care decisions [46].

Author Contributions

Conception/design: Siew Tzuh Tang, Chen Hsiu Chen, Fur‐Hsing Wen, Ming‐Mo Hou, Chia‐Hsun Hsieh, Wen‐Chi Chou, Jen‐Shi Chen, Wen‐Cheng Chang

Provision of study material or patients: Ming‐Mo Hou, Chia‐Hsun Hsieh, Wen‐Chi Chou, Jen‐Shi Chen, Wen‐Cheng Chang

Collection and/or assembly of data: Siew Tzuh Tang, Chen Hsiu Chen, Ming‐Mo Hou, Chia‐Hsun Hsieh, Wen‐Chi Chou, Jen‐Shi Chen, Wen‐Cheng Chang

Data analysis and interpretation: Siew Tzuh Tang, Chen Hsiu Chen, Fur‐Hsing Wen

Manuscript writing: Siew Tzuh Tang, Chen Hsiu Chen

Final approval of manuscript: Siew Tzuh Tang, Chen Hsiu Chen, Fur‐Hsing Wen, Ming‐Mo Hou, Chia‐Hsun Hsieh, Wen‐Chi Chou, Jen‐Shi Chen, Wen‐Cheng Chang

Disclosures

Jen‐Shi Chen: Ono Pharmaceutical Co., Ltd., Novartis Taiwan (C/A). The other authors indicated no financial relationships.

(C/A) Consulting/advisory relationship; (RF) Research funding; (E) Employment; (ET) Expert testimony; (H) Honoraria received; (OI) Ownership interests; (IP) Intellectual property rights/inventor/patent holder; (SAB) Scientific advisory board