INTRODUCTION

The relationship between a patient’s life expectancy and future treatment, screening, and diagnostic decisions is an important consideration during individual care planning. Estimating the potential remaining years of life is particularly relevant for those with a limited estimated life expectancy, such as older people with frailty.1,2,3,4 Many screening and treatment guidelines differ, based on life expectancy.5,6 For example, less intensive diabetic treatment targets are recommended for people with a life expectancy of <10 years.5 Discussing life expectancy allows patients to make informed choices about their care.

Despite this, healthcare professionals may be reluctant to initiate discussions about life expectancy, as they may fear patient reactions to the topic, feel they are taking away hope, or want to respect cultural values of individual patients.7,8,9 Several studies have explored patient preferences for discussing life expectancy with widely varying results. For example, two studies among patients with cancer reported a willingness to discuss life expectancy of 19%10 and 81%,11 respectively. Previously published systematic reviews have primarily focused on patients with cancer.12,13,14

The aim of our systematic review was to provide a comprehensive overview of the literature on patient preferences for discussing life expectancy and to identify predictors for willingness to discuss life expectancy.

METHODS

We conducted a systematic review according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines.15 Our population of interest was individuals aged ≥18 years. We included studies regardless of patient group. We were interested in studies which reported on patient preferences, views, and attitudes toward discussing life expectancy. Our outcome of interest was the proportion of patients willing to discuss life expectancy with a healthcare provider. Studies were included regardless of their aim; that is, they only had to report a proportion of patients being willing to discuss life expectancy. Further, all study designs (that is, quantitative and qualitative) were included. Studies that solely considered the perspective of those other than patients, for example relatives and family, were excluded.

Search Strategy

We searched the databases PubMed, Cochrane Library, CINAHL, Embase, MEDLINE, and PsycInfo from inception to February 17, 2021. We combined keywords for the following concepts: (1) “discussion, communication, decision making”; (2) “life expectancy”; and (3) “preferences, opinion, view, perspective, attitude, belief.” Non-English studies were translated and assessed for eligibility.16 The full search strategies are outlined in Appendix 1. Reference lists of eligible studies were also manually searched while a gray literature search was preformed using Google Scholar and the Turning Research Into Practice database.

Screening

Titles and abstracts were first screened independently by two authors (E. B. and W. T./C. L.). Full texts of relevant studies from the first step were then screened by two authors (E. B. and W. T./C. L.). Any disagreements were resolved via group discussion among these three authors.

Critical Appraisal

Two tools were used for critical appraisal of eligible studies. Quantitative survey/questionnaire studies were assessed using the checklist for Critical Appraisal of a Survey by Center for Evidence-Based Management.17 Interview studies with a qualitative study design were assessed using the Qualitative Checklist by Critical Appraisal Skills Programme (CASP).18 The full appraisals are reported in Appendix 2. Critical appraisal was performed by one author (E. B.). Uncertainties were resolved via group discussion with two other authors (W. T., C. L.).

Data Extraction

Two authors (E. B. and C. L.) extracted data on study characteristics and the outcome of interest from the eligible studies. For this review, every positive degree of willingness to discuss life expectancy was included. For example, if the question was “If you had advanced chronic illness, would you like to be informed about whether it is likely that you may not live more than a few months?”, responses such as “Yes, but only when I ask” and “Yes, it is the doctors duty to tell me even if I don’t ask” were both extracted as positive degrees of willingness to discuss.

Data Analysis

We attempted to conduct meta-analyses for all studies and pre-specified subgroups (in order to evaluate predictors of willingness to discuss life expectancy), planning to report pooled proportions through RStudios.19 We were interested in the following sub-groups: patient type, geographical location, age profile, and study design. We encountered high heterogeneity for analyses of both the overall group of studies and sub-groups (I2>90%); we therefore instead reported descriptive statistics only and conducted a narrative synthesis of the results. Studies were synthesized by comparing and interpreting extracted data to group similar studies together based on the sub-groups identified above.

RESULTS

Our search returned a total of 6,762 titles and abstracts, leaving 4,312 studies eligible for screening after removal of duplicates. We screened 224 full-text studies. A total of 41 studies (n=27,570 study participants) met the eligibility criteria. The PRISMA flowchart is presented in Figure 1. Three pairs of included studies20,21,22,23,24,25 reported on the same study data but only one dataset from each was included in the analyses. Characteristics of all included studies are presented in Table 1; however, a subset of studies20,23,25,26,27,28,29,30,31,32,33,34 were not included in calculating the range of proportions, median, or interquartile range (IQR) due to either not reporting a numeric proportion,26,27,31,32 being conducted among members of the public with no specific diagnosis,28,29,33 or reporting on a dataset already included.20,23,25 Two studies were excluded from the age analysis due to reporting age as an interval and not as mean/median age.30,34 One study was excluded from the sub-analysis of advanced cancer/non-advanced cancer35 (see details in Appendix 3).

Figure 1
figure 1

PRISMA flowchart of review process.15

Table 1 Descriptive Table Outlining Study Characteristics and Proportion Willing to Discuss Life Expectancy

Characteristics of Included Studies

Two different study types were included: quantitative survey/questionnaire studies (n=27),11,22,23,24,25,28,29,30,33,35,36,37,38,39,40,41,42,43,44,45,46,47,48,49,50,51,52 including one baseline survey from a prospective randomized clinical trial,51 and qualitative interview studies (n=14),10,20,21,26,27,31,32,34,53,54,55,56,57,58 including one baseline interview from a prospective cohort study.34 Studies were predominantly from North America,22,23,30,32,34,36,40,42,46,51,52,56,57,58 Australia,10,11,24,25,26,29,39,45,53 and Western Europe,20,21,27,28,31,33,35,37,41,48,55 while five studies were from East Asia38,44,47,49,54 and two studies were from West Asia.43,50 A total of eight participant groups were represented: patients with cancer,10,11,20,21,24,25,26,31,32,34,35,38,39,40,43,44,45,48,49,50,52,54,55 surgical outpatients,37 patients with congenital heart disease,36,41,51,58 nephrology outpatients,42 older adults,22,23,27,46,56,57 patients with Parkinson’s disease,30 inpatients with mixed advanced diseases,47,53 and members of the public with no specific diagnosis.28,29,33 The age profile was predominantly older patients, while five studies reported on a study population with a mean/median age below 50 years.36,41,43,45,50 A table detailing characteristics of the included studies can be found in Appendix 4.

Willingness to Discuss Life Expectancy

The proportions of patients willing to discuss life expectancy in each study are presented in Table 1 and Figure 2. Four studies reported a non-numeric proportion of participants willing to discuss life expectancy: Mathie et al. (2012)27 and Deschepper et al. (2008)31 reported the proportion willing to discuss life expectancy as “minority”, Sherman et al. (2018)32 reported the proportion as “varied” and Clayton et al. (2005)26 reported it as “most”. The remaining studies reported numeric proportions, which varied from 19 to 100% with a median of 61% (IQR 50–73). In total, 77% of numeric studies (n=24) reported a proportion willing to discuss life expectancy above 50%.

Figure 2
figure 2

Forest plot with all studies eligible for analyses in RStudios,19 depicting the overall heterogeneous distribution. The forest plot has a range from 0 to 1, correlating to proportions of willingness to discuss life expectancy between 0 and 100%. Studies to the left of the 0.5 axis (n=7) reported a willingness to discuss life expectancy below 50%, while studies to the right of the 0.5 axis (n=2) reported a willingness to discuss life expectancy 50% or above. *Hoesseini 2020 reported the proportion 100%, which equals a standard error of 0. To graphicly represent the study in this forest plot a simulated standard error of 1.0E-8 has been plotted.

Patient Type

Patients with cancer were the most frequently studied group (n=23). Among patients with cancer, the proportion willing to discuss life expectancy ranged from 19 to 100% (median 61%, IQR 51–81)10,11,21,24,26,31,32,34,35,38,39,40,43,44,45,48,49,50,52,54,55 including three non-numeric studies: “minority”,31 “varied”,32 and “most”26 (Fig. 3). Around one-fifth (22%) of the studies reported a proportion below 50%. In studies among patients with non-advanced cancer (n=10),24,38,39,43,44,45,49,50,54,55 20% of studies (n=2) reported a willingness below 50%, while in studies among patients with advanced cancer (n=10)10,11,21,26,31,32,34,40,48,52 two studies (20%) similarly reported a willingness below 50% (Appendix 5, eFigure 1). However, the advanced cancer group had a wider range (19–94%, median 73%, IQR 37–85), including the three non-numeric studies,26,31,32 compared to the non-advanced cancer group (30–100%, median 59%, IQR 51–61).

Figure 3
figure 3

Forest plots over sub-analysis based on patient type in RStudios.19 The forest plot has a range from 0 to 1, correlating to proportions of willingness to discuss life expectancy between 0 and 100%. Studies among patients with cancer (n=17) has four studies on the left of the 0.5 axis, meaning the willingness to discuss was below 50%. Studies among patients with heart disease (n=4) and older patients (n=4) each had one study to the left of the axis. Among studies with inpatients with mixed advanced disease (n=2), all studies were to the left or on the 0.5 axis. Studies conducted among surgical outpatients (n=1), patients with Parkinson’s disease (n=1), and nephrology outpatients (n=1) reported all studies to the right of the 0.5 axis. *Hoesseini 2020 reported the proportion 100%, which equals a standard error of 0. To graphicly represent the study in this forest plot a simulated standard error of 1.0E-8 has been plotted.

Other studied groups included patients with heart disease (n=4), older patients (n=5), surgical outpatients (n=1), Parkinson’s disease patients (n=1), nephrology clinic outpatients (n=1), and inpatients with mixed diseases (n=2). The proportion willing to discuss life expectancy among patients with heart disease (n=4) ranged from 35 to 71% (median 64%, IQR 48–69).41,36,51,58 Among older patients (n=5), the range was 41–68% (median 66%, IQR 53–67)46,22,27,56,57 including one non-numeric study where a “minority” was willing to discuss.27 Among surgical outpatients (n=1), the proportion was reported as 58%.37 The study conducted among patients with Parkinson’s disease (n=1) reported 87%30 of patients were willing to discuss life expectancy. Among first time outpatients at a nephrology clinic (n=1), a similar high proportion was reported: 97%.42 Among inpatients with mixed diseases (n=2), all studies reported a willingness to discuss at 50% or below: ranging from 23 to 50%.47,53

Geographical Location

The proportion willing to discuss life expectancy ranged from 21 to 97% (median 67%, IQR 63–80, Appendix 5, eFigure 2) 40,46,32,36,56,22,51,57,58,52,34,30,42 in North America (n=13) including one non-numeric proportion where the willingness “varied”;32 from 19 to 81% (median 59%, IQR 23–66)10,53,45,39,24,11,26 in Australia (n=8) including one non-numeric proportion where “most” were willing to discuss;26 from 35 to 100% (median 73%, IQR 52–89)21,27,31,35,37,41,48,55 in Western Europe (n=7) including two non-numeric proportions where a “minority” were willing to discuss;27,31 from 30 to 58% (median 50%, IQR 45–51)54,44,38,47,49 in East Asia (n=5); and 60 to 61% in West Asia (n=2).43,50 All locations had predominantly positive preferences for discussing life expectancy except East Asia, where almost half (40%) of the studies reported a proportion lower than 50%.

Age Profile

Among studies with a mean/median age of <50 years (n=5), the proportion of patients willing to discuss life expectancy ranged from 35 to 61% (median 60%, IQR 57–61, Appendix 5, eFigure 3).41,43,45,36,50 For studies with a mean/median age of 50–64 years (n=14), the proportion ranged from 19 to 97% (median 61%, IQR 45–87),10,40,54,38,32,44,49,39,58,52,11,21,42,35 including one non-numeric proportion where the willingness to discuss “varied”.32 For studies with an age profile of ≥65 years (n=13), the proportion willing to discuss life expectancy ranged from 23 to 100% (median 65%, IQR 51–66),22,24,26,27,37,46,47,48,51,53,55,56,57 including two non-numeric studies: in one study “most” were willing to discuss life expectancy 26 and in another a “minority” were.27 All age subgroups had similar trends, with one third or less of studies reporting a proportion below 50%. The studies in the lowest age subgroup (<50 years) and highest age subgroup (≥65 years), reported a proportion willing to discuss below 50% in around one-fifth (20%,17%) of the studies.41,46,53 One-third of the studies in the middle age group (50–64 years) reported a proportion below 50%.10,38,40,54

Study Design

Lastly, we analyzed studies based on study design (Appendix 5, eFigure 4). Studies with a quantitative study design (n=22), including one baseline survey from a prospective randomized clinical trail,51 reported the proportion of willingness to discuss life expectancy ranging from 21 to 97% (median 61%, IQR 51–66).11,22,24,30,35,36,37,38,39,40,41,42,43,44,45,46,47,48,49,50,51,52 The majority of the quantitative studies reported a proportion willing to discuss life expectancy above 50%. Among the qualitative studies (n=13), including one baseline interview from a prospective cohort study34 and four non-numeric studies (reporting different degrees of willingness to discuss: “minority”,27,31 ”varied”,32 and “most”26), the proportion willing to discuss life expectancy ranged from 19 to 100% (median 68%, IQR 30–73),10,21,26,27,31,32,34,53,54,55,56,57,58 with two-thirds reporting a proportion above 50% willingness to discuss life expectancy.

DISCUSSION

Willingness to discuss life expectancy varied widely. Most patients preferred to discuss life expectancy, with 77% of studies reporting the majority would be willing to discuss life expectancy. The variability in willingness to discuss life expectancy across studies highlights the complex interplay of the many factors (age group, patient type, geographical location, disease stage) that shape patient attitudes on this topic. Unfortunately, for providers, this complexity and variability, even among relatively homogenous patient populations, make it challenging to definitively identify predictors and patterns.

There may be several reasons why there was wide variability. Asian countries reported less willingness to discuss life expectancy among participants, with almost half (40%) studies reporting a willingness lower than 50%. Previous systematic reviews on preferences of East Asian patients with cancer found a similar outcome, with 30% willingness to discuss life expectancy.12,14 Being unaware of death is considered an important aspect of a good death, which could possibly explain this cultural trend.59,60 Participants from studies conducted in North America, Australia, West Asia, and Western Europe were generally positive toward discussing life expectancy, with most studies reporting proportions above 50%. There was no clear pattern with respect to age or patient type, despite wide variability. Another challenge in comparing proportions across studies is that the question and response options differed between studies (Appendix 6).

Previous reviews focused on advanced cancer and severe illness reported that patients generally tend to be less willing to discuss life expectancy as their disease progresses.14,61 In our review, however, we did not identify a clear pattern in the advanced cancer group, but we did find that all studies conducted among inpatients with mixed advanced diseases found that ≤50% of patients were willing to discuss their life expectancy. Like advanced cancer patients, we found large variability in studies of patients with non-advanced cancer. The large variability found in the two cancer groups is possibly due to patients with cancer being the most studied patient group. The general perception that a cancer diagnosis is a life-limiting diagnosis, no matter the disease progression,62,63 might explain the similar findings in these two groups.

Both quantitative and qualitative studies were included in this review. Previous literature has shown that the quality of data on life expectancy is influenced by the study method,64 with the willingness to disclose sensitive information possibly being lower when data is collected during face-to-face interviews and higher when using self-administrated surveys. This might result in an underestimation of otherwise positive attitudes to discussing life expectancy in qualitative studies. We found a slightly lower median willingness to discuss among qualitative studies (67%) than among quantitative studies (81%). It is also possible that physician type, physician characteristics, or the clinical setting may influence patient attitudes toward discussing life expectancy. We did not extract data on differences in attitudes related to who the patient would be having a discussion with. It may be that patients would feel more comfortable communicating about life expectancy with their general practitioner with whom they have a long-lasting and trusting relationship.65,66

Limitations

This review is limited by several factors. First, in the eligible studies, willingness to discuss life expectancy was explored under the conditions of a research study. Thus, the attitudes expressed may not reflect how patients would feel about discussing life expectancy in the context of actual clinical care (e.g., if the consideration was processed over time and considered with a trusted healthcare professional). Further, factors such as physician characteristics and the setting, which we did not capture, may affect how patients feel about these discussions. Second, while we report individual answer categories, we chose to combine every positive degree of willingness to discuss life expectancy to create one proportion, for example preferences such as “Yes, in principle always” and “Yes, only if I ask.” This may have led to an overestimation of the proportion in studies with multiple positive response options (n=13). Third, we included studies regardless of sample size, possibly introducing greater variance within the proportions. However, as we also calculated ranges excluding studies with less than 100 patients and found no significant changes, we kept all studies in the analyses (Appendix 7). Fourth, including qualitative studies likely introduced bias, because qualitative study samples are usually purposefully selected.67 Finally, included studies were very diverse in terms of framing of the question, study methods, and disease types, which resulted in high heterogeneity and made it difficult to compare results across studies. Specifically, the question about discussing life expectancy was framed in very different ways: some studies asked directly, other studies set up a hypothetical patient scenario, while studies also included different response options from one another. This made comparison between studies difficult (Appendix 6).

Clinical Implications

We could not identify any clear trends or predictors for patient preferences. As such, our findings stress the importance of considering every patient individually in terms of initiating discussions about life expectancy. While life expectancy is important to discuss for screening and treatment decisions in those with limited life expectancy, achieving a shared understanding around what and how much patients would like to know and discuss appears to be a good starting point during consultations that may deal with life expectancy. Despite its importance, it appears that clinicians cannot assume a willingness to discuss life expectancy based on specific characteristics, as people may have vastly different preferences regardless of similar, age, disease status, and culture. Preferences may change with age as health status changes and/or diseases progress. Therefore, clinicians should consider re-evaluating preferences around discussing life expectancy periodically and consider bringing it up when there may be a relevant reason to discuss it (e.g., at nursing home admission). Such discussions should incorporate the entire patient context (culture, health status, care preferences).

Implications for Future Research and Clinical Practice

This review highlights that preferences for discussing life expectancy cannot be easily summarized or generalized into precise guidance for clinical practice. Since the preferences for discussing life expectancy were generally positive but with large variations, future studies should explore mechanisms for determining patient preferences around how, if, and when to discuss life expectancy. Furthermore, research is needed with a focus on evaluating better integration of discussions into routine clinical practice, perhaps tools for electronic records that prompt discussion around preferences for discussing life expectancy or prompt re-evaluation of previous discussions in the transition of patients through health statuses. Similarly, more research is needed among patient types other than patients with cancer. Discussing life expectancy is important among all patients with limited life expectancy, despite the literature mainly focusing on patients with cancer.

CONCLUSION

The heterogeneity documented in this review underscores the highly individual and complex nature in which patients deal with life expectancy and thus the need to consider each patient individually. While most patients are positive toward discussing their life expectancy, there was variability across studies. Reasons for the wide variability may to some extent be due to geographical location, age, patient type, framing of the discussion, or even physician relationship. Variability and heterogeneity make it challenging to identify clear predictors of willingness to discuss life expectancy. Despite including all patient types, more than half of all studies were conducted among patients with cancer, highlighting the need for more studies examining life expectancy preferences among other patient groups. Future research should explore mechanisms for practically determining patient preferences for how, if, and when to discuss life expectancy, to be incorporated into existing guidelines, and ultimately into routine clinical practice.