Article Text

Download PDFPDF

Discordance between patients' stated values and treatment preferences for end-of-life care: results of a multicentre survey
  1. Daren K Heyland1,2,3,
  2. Rebecca Heyland3,
  3. Peter Dodek4,
  4. John J You5,
  5. Tasnim Sinuff6,7,
  6. Tim Hiebert8,
  7. Xuran Jiang3 and
  8. Andrew G Day3
  9. for the ACCEPT Study Team and the Canadian Researchers at the End of Life Network (CARENET)
  1. 1Department of Critical Care Medicine, Kingston General Hospital, Kingston, Ontario, Canada
  2. 2Department of Public Health Sciences, Queen's University, Kingston, Ontario, Canada
  3. 3Clinical Evaluation Research Unit, Kingston General Hospital, Kingston, Ontario, Canada
  4. 4Centre for Health Evaluation and Outcome Sciences and Division of Critical Care Medicine, St. Paul's Hospital and University of British Columbia, Vancouver, British Columbia, Canada
  5. 5Departments of Medicine, and of Clinical Epidemiology & Biostatistics, McMaster University, Hamilton, Ontario, Canada
  6. 6Department of Critical Care Medicine, Sunnybrook Hospital and Sunnybrook Research Institute, Toronto, Ontario, Canada
  7. 7Interdepartmental Division of Critical Care Medicine and Department of Medicine, University of Toronto, Toronto, Ontario, Canada
  8. 8Winnipeg Regional Health Authority Palliative Care Program, Winnipeg, Manitoba, Canada
  1. Correspondence to Dr Daren K Heyland, Department of Critical Care Medicine, Kingston General Hospital, Angada 4, Kingston, Ontario, Canada K7L 2V7; dkh2{at}


Background Medical orders for the use of life-supports should be informed by patients' values and treatment preferences. The purpose of this study was to explore the internal consistency of patients' (or their family members') stated values, and the relationship between these values and expressed preferences.

Methods We conducted a prospective study in 12 acute care hospitals in Canada. We administered a questionnaire to elderly patients and their family members about their values related to end-of-life (EOL) care, treatment preferences and decisional conflict.

Results Of 513 patients and 366 family members approached, 278 patients (54%) and 225 family members (61%) consented to participate. Participants' most important stated values were to be comfortable and suffer as little as possible, to have more time with family, to avoid being attached to machines and tubes and that death not be prolonged. The least important stated value was that life be preserved. Based on prespecified expected associations between the various values measured, there were inconsistencies in participants' expressed value statements. With few exceptions, participants' expressed values were not associated with expected corresponding treatment preferences. Of the 109 (40%) patients and 95 (42%) family members who had made decisions about use of life-supports, 68 (56%) patients and 60 (59%) family members had decisional conflict.

Conclusions Decision-making regarding medical treatments at the EOL is inadequate. To reduce decisional conflict, patients and their families need more support to clarify their values and ensure that their preferences are grounded in adequate understanding of their illness and treatment options.

Trial registration number NCT01362855.

  • Communication
  • Ethics
  • Clinical decisions
  • Hospital care

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.


For the past several decades, end-of-life (EOL) decision-making in Western cultures has been rooted in the ethical principle of autonomy whereby patients have the right to receive care that is consistent with their values and treatment preferences. In 2014, the Institute of Medicine's report on EOL care in America1 re-emphasised that “patients nearing the end of life ought to receive treatments that match their preferences and goals of care”, and recommended that patients complete advance directives to achieve this goal. However, this report and other publications note the lack of success of advance directives in influencing subsequent clinical care.1–3 Furthermore, several studies document the instability of some patients' preferences over time,4 that patients are inadequately informed about treatment options at the EOL5 and that clinicians often do not abide by the written or oral instructions made in advance.6 ,7 We postulate that patients are ill-prepared to make decisions about the kinds of medical treatments they would want at the EOL. Thus, when patients have a serious illness, clinicians may be uncertain about whether the patient's prior directives are based on a truly informed decision and are grounded in his or her authentic values.

We advocate that our focus should be on improving the quantity and quality of advance care planning (ACP) as a way of preparing patients and their substitute decisionmakers to be ‘decisionally ready’ to meet with clinicians near or at the EOL to make decisions about the use of life-supporting treatments.8 We define ACP as a process where people plan for a time when they cannot make decisions for themselves.9 This process includes reflection, deliberation, determination of a person's authentic values, informed preferences for treatments at the EOL and the communication of these ideas. By values, we mean an expression of a person's overarching philosophies or most important priorities in life (such as maximising quality of life or time spent with family). Wishes and preferences are used interchangeably and refer to specific preferred options for treatments (such as use of resuscitative treatments or not) or preferred health states (conditions that would be acceptable or unacceptable). These preferences are generally expressed outside the clinical context and are not to be misconstrued as a medical decision. Although patients may be able to identify what is important to them in their life and how they do and do not want to live, it is hard for patients and families (who have not had medical training) to translate their views into specific medical treatments and procedures, especially as they often do not understand the ramifications or the likely outcomes of those treatments. A medical decision requires a conversation with a healthcare professional and consideration as to whether the treatment options are clinically indicated.9 If decision-making around treatment options is not carried out optimally, decisional conflict, defined as a state of uncertainty about a specific action, may result.10

An understanding of where patients and families experience difficulties in making decisions about treatment options near the EOL is important to guide future quality improvement initiatives. The purpose of this study was to explore the internal consistency of patients' (or their family members') stated values, the relationship between these values and expressed preferences, to ascertain whether there was any decisional conflict and whether decision conflict related to any specific preferences. We hypothesised that a significant proportion of patients and family members would be unable to articulate their values and preferences, that their expressed values would be internally inconsistent and not correlate with expressed preferences, and that some patients would experience decisional conflict. For example, we hypothesised that patients and families would express that living as long as possible and avoiding pain and discomfort would be equally important, that neither would predict the preference for cardiopulmonary resuscitation (CPR) and that the preference for CPR would be associated with increased decisional conflict.


We conducted a multicentre prospective audit of communication, planning and documentation at the EOL in a convenience sample of 12 teaching hospitals in Canada. A prior report from this audit outlines our methods in detail.11 In brief, hospitalised patients were enrolled in this study if they had advanced pulmonary, cardiac or liver disease, metastatic cancer (see online supplementary appendix A for details), or if they were 80 years of age or older and admitted to hospital, but not to the intensive care unit, from the community for an acute medical or surgical condition. If none of the above criteria were met, any patient whose death within the next 6 months would not surprise any member of their healthcare team was also eligible.12

We approached consecutive, eligible patients and their family members for their consent to participate. Patients unable to communicate due to language or for cognitive reasons were excluded, but if a family member was available, he or she was approached independently. Enrolled study patients were asked to identify an adult family member who knew them the best (inclusive of partners, significant others and/or close friends) and who had visited the patient in hospital at least once. We first approached the patient and family member between 48 and 120 hours after admission of the patient to allow for abatement of symptoms present at the time of admission so that the patient and family member could participate in an interview.

On enrolment, the research assistant conducted separate face-to-face interviews with patients and family members. Details of the questionnaire used in these interviews are described elsewhere.13 In brief, questions addressed the patient's and family member's engagement in EOL decisions during the current hospitalisation. To assess the underlying values of respondents, we used the End of Life Values Scale,14 a multi-item questionnaire that asks the respondent to rate the level of importance (1=not important; 10=extremely important) of each item in a list of common EOL values. This scale has been shown to have good internal consistency and predictive validity in long-term care settings.14

To elicit preferences for use of life-sustaining treatments, we used a taxonomy describing different levels of the use or non-use of life-sustaining treatments (eTable 1). This taxonomy was developed with inputs from medical experts and has been used extensively in our prior research.2 ,11 ,13 Patients and family members were asked to express their own sense of underlying values and preferences about the patient's care, not as a proxy for the other participant.

For those participants who responded ‘yes' to “Was a decision made about whether to use or not to use life-sustaining treatments in the event of a life-threatening illness during this hospital stay?”, we also administered a set of additional questions to assess their decisional conflict. To get a more comprehensive understanding of the nature of their decisional conflict, we expanded the existing validated four-item decision conflict scale15 to seven items (table 1). New items were generated by consensus from the investigator team, and had face and content validity. For each item, a ‘yes' response was awarded one point. As suggested by the developers of the Decisional Conflict Scale,13 respondents were considered ‘decisionally ready’ if they scored 7 points (maximum score) and ‘decisionally conflicted’ if they scored <7.

Table 1

Decisional conflict regarding patient's health options

Statistical analysis

We attempted to obtain a representative sample of patients and families from each participating site (minimum of 20). Thus, the total sample size was a function of the number of participating hospitals and their success in enrolling a representative sample of patients and families in this quality improvement initiative.

Demographics of patients and their family members were described as means and ranges for continuous variables, and as counts and percentages for categorical variables. For skewed data, such as the EOL Values Scale, we report median and IQRs of the ratings. To assess the internal consistency of respondents' stated values, we calculated the Pearson correlation coefficients among the various value statements in the EOL values scale. Since the individual items in the EOL Values Scale are associated with prolongation or shortening of life, a priori we did not expect all items to be significantly correlated. Rather, based on the content and directionality of the individual items, two members of the study team (DKH and RH) developed the expected relationships among these individual value statements (shown in table 2), and the observed and expected relationships were defined.

Table 2

Correlations between values about the use of life-supporting treatments

Box plots have been used to describe the distribution of the importance of stated values by the preference for life-supporting treatment. The concordance between respondents' stated values and the preferences for life-supporting treatments (ranked as 1-comfort care only, 2-some life-sustaining treatment and 3-aggressive life-sustaining treatment including CPR) was measured by Kendall's τ-b statistic. Kendall's τ-b is the excess concordance (C-(1-C)) where C is the probability that an individual who puts a higher importance on a given value preferred more aggressive care than someone who put a lower importance on the same value. Kendall's τ-b ranges from 1 to −1 where 1 indicates perfect concordance, −1 indicates perfect discordance and 0 indicates no association. Decisional conflicts regarding patient treatment preferences were described by counts and percentages or means and SDs. A priori we postulated that decisional conflict would be higher in patients who preferred resuscitation measures. The Mann-Whitney U test was used to test the difference in the decisional conflict score between patients who preferred comfort measures only versus those who preferred the use of resuscitative measures.

This study was approved by the Research Ethics Boards of all participating institutions, and all participants provided written informed consent.


We approached 513 eligible patients and 366 eligible family members; 278 patients and 225 family members consented yielding an enrolment rate of 54% and 61%, respectively (eFigure 1). The average age of patients was 80 years and 37% lived alone (eTable 2). Average age of family members was 61 years; most were children of the patient (56%; eTable 3). Most patients and family members were Caucasians who spoke either English or French. Of note, 83% of enrolled patients had a medical order for the use (or non-use) of life-sustaining treatments found on the medical record on the day of enrolment.

Underlying values

The most important stated values were to be comfortable and suffer as little as possible, to have more time with family, to avoid being attached to machines and tubes and that death not be prolonged. The least important stated value was that life be preserved (table 3). For both patients and families, rates of ‘unsure or undecided’ were <10% for all the value statements.

Table 3

Importance of values and beliefs related to end-of-life care

Table 2 provides the correlation among values about the use of life-sustaining treatments. Of the 18 correlations among values that were expected a priori among patients, only 7 (39%) were statistically significant in the expected direction. Among families, 5 (28%) were statistically significant. All positive correlations were weak (0.14) to moderate (0.60) in terms of the magnitude of the association. For example, we had postulated that ‘living as long as possible’ would positively correlate with the ‘belief that life should be preserved.’ We did observe a significant positive correlation but the coefficient was only 0.35 (p<0.0001). In contrast, we expected that ‘living as long as possible’ would be negatively correlated with ‘be comfortable and suffer as little as possible’, ‘avoid being attached to machines and tubes', ‘death is not prolonged’ and the ‘belief that nature should be allowed to take its course’. However, we did not find a significant negative correlation in any of these instances. In fact, from the patient's perspective, ‘be comfortable and suffer as little as possible’ was correlated only weakly with ‘live as long as possible’ (correlation coefficient, 0.14, p=0.03).

Expressed preferences

Most patients preferred either comfort care (31%) or a mix of comfort and full medical care that does not include resuscitation (31%). Only 12% preferred aggressive medical care, including resuscitation (see eTable 1). Similarly, family members most commonly preferred either comfort care (35%) or a mix of comfort and full medical care that does not include resuscitation (28%) whereas only 15% preferred aggressive medical care, including resuscitation for their loved ones (see eTable 1). Eight per cent of patients and family members were unsure of their treatment preferences.

Relationship between values and preferences

With few exceptions, ratings of importance of the various values did not discriminate between different preferences for the use of life-support measures (figure 1 and eFigure 2). For example, all respondents, regardless as to whether they preferred CPR or not, rated ‘being kept comfortable and suffering as little as possible’ an average of 10 of 10 on the importance scale. Similarly, ratings of importance were not different among groups of patients that preferred varying levels of life-sustaining treatments for the values ‘avoid being attached to machines and tubes’, ‘my death not be prolonged’ and that ‘nature be allowed to takes its course’. In contrast, there were significant differences in ratings of importance for the values related to living as long as possible and that life be preserved (figure 1). The Kendall's τ-b was consistently below 0.3 indicating only a weak association between stated values and expressed preferences (figure 1 and eFigure 2).

Figure 1

The relationship between importance of values and treatment preferences—patients' perspectives. The box plots show the distribution of importance of values for each treatment preference. Kendall's τ-b is the excess concordance (C-(1-C)) where C is the probability that an individual who puts a higher importance on a given value preferred more aggressive care than someone who put a lower importance on the same value. Kendall's τ-b ranges from 1 to −1 where 1 indicates perfect concordance, −1 indicates perfect discordance and 0 indicates no association. For example, Kendall's τ-b would be 1 if everyone who preferred aggressive life-sustaining treatment provided a higher importance rating to living as long as possible than anyone else, and if everyone who preferred comfort care only gave a lower importance rating to living as long as possible than anyone else. Conversely, Kendall's τ-b would be −1 if everyone who preferred comfort care only rated the importance to avoid being attached to machines and tubes higher than everyone else, and everyone who preferred aggressive life sustaining treatment rated the importance to avoid being attached to machines and tubes lower than everyone else. Treatment preference: (1) Doctors will be focused on my comfort, alleviating suffering, and not on being kept alive by artificial means or heroic measures such as trying to prolong my life with CPR and other life-sustaining technologies; (2) A mix of the options (eg, try to fix problems but if not getting better switch to focusing only on my comfort even if it hastens death); (3) Full medical care but in the event my heart stops or my breathing stops, no CPR; (4) Aggressive use of heroic measures and artificial life-sustaining treatments, including CPR, to keep me alive at all costs. Importance of values: 1=‘Not at all important’; 10=‘Very important’. CPR, cardiopulmonary resuscitation.

Decisional conflict

Of all the participants, 109 (40%) patients and 95 (42%) family members stated that they had made a decision regarding the use of life-supporting treatments during the hospitalisation. Of these 204, 85% responded that they had made the best decision possible whereas 128 (57%) expressed decisional conflict (score of <7): 68 (56%) patients and 60 (59%) family members (table 1). The main source of decisional conflict was a lack of knowledge of the disease and treatment options. Patients who preferred aggressive medical treatments, including CPR, had a greater decisional conflict compared to patients who preferred comfort measures only (6.0 vs 4.8, p=0.01).


In this multicentre audit of communication and decision-making practices in 12 hospitals in Canada, we explored the relationship between patients' (and family members') stated values about EOL care and their preferences for use or non-use of life-supporting treatments. We observed that more than 90% of respondents were able to rate the importance of each value in the EOL values scale, but many value statements seemed to conflict with each other and very few of the values were associated as expected with the expressed preference for use of life-sustaining treatments. Perhaps the most striking example of this was that participants stated that ‘to be kept comfortable’ was their most important value but 12–15% still preferred all life-sustaining measures, including CPR. Only 40% of the participants recognised that a decision about their goals of care had been achieved during this hospitalisation despite the fact that the majority of these patients had medical orders for the use (or non-use) of life-sustaining treatments on their charts.11 More than half of those who recognised that a decision had been made still had some decisional conflict. Importantly, this decisional conflict was higher in those patients who desired CPR compared to patients who preferred less aggressive care. These findings have important implications for clinical practice and on how we train clinicians to have these EOL conversations.

How can clinical decision-making be improved so that care plans are consistent with patients' most important values and also associated with minimal decisional conflict? Our findings raise questions about some patients' ability to recognise the conflict between different values and the trade-offs inherent in making EOL treatment decisions, such as the trade-off between quality and quantity of life. In isolation, participants found most of the values to be very important. They want to be kept comfortable as well as to have their life prolonged so they can have more time with family. To move forward, patients will need greater help to clarify their underlying values, to appreciate that trade-offs may be required and in situations where trade-offs are required (eg, life prolongation may require sacrificing comfort or enduring greater physical suffering), they will need help in prioritising their values. Decision aids that use techniques to elicit patient values about the consequences of a treatment decision and explicitly ask patients to rate how much those consequences matter or to rank them may be more effective than tools that do not help patients to resolve conflicting values or trade-offs.16 There are several tools that contain values clarification techniques such as the Making Your Wishes Known,17 PREPARE website18 and the Graphic Values History Tool (Personal Communication: P Allatt). These tools may have some clinical utility in better preparing patients for clinical decision-making. Given that the majority of older patients will lack competency at the point of EOL medical decision-making,19 it is vitally important that these tools be used before the onset of acute illness and that the values clarification process involves future surrogate decisionmakers.

The current study found that most decisional conflicts relate to inadequate knowledge of risk, benefits and alternatives of various treatment options (table 1), and the majority of decisional conflict resides with patients seeking aggressive care at the EOL. In a prior study, Heyland et al5 documented that this older hospitalised patient population is poorly informed about the nature, associated treatments and outcomes of CPR. Rather than automatically accepting patients' stated wishes for resuscitation, perhaps patients who express a preference for CPR should watch a CPR video decision aid or use a similar paper-based tool to confirm their treatment preferences.20 ,21 There is emerging evidence that tools to promote shared decision-making in the context of serious illness reduce decisional conflict and can improve perceived quality of care.22–24 Integration of these tools into the healthcare system before and at the point of decision-making is vital.

It is also apparent from our findings that many patients do not recognise that decisions have been made on their behalf. Although more can be done to better prepare patients for future decision-making encounters,8 healthcare professionals must also take responsibility for the quality of the clinical decision-making encounter. Healthcare professionals must inform their high-risk patients about their prognosis, ask their patients about their values, fears and concerns, answer outstanding questions and elicit an informed preference for treatments at the EOL.25 They should not accept ‘verbatim’ any expression of a treatment preference as the gold standard of ‘what’ is in the patient's best interest. Rather, they should carefully explore the trade-offs inherent in the patient's wishes and medical condition. To facilitate awareness and capacitation of these skills, a number of tools have been developed to help aid healthcare professionals to improve these interactions with their patients.26–28 Moreover, simply copying the medical order from prior documents, without (re)addressing the patient's underlying values and preferences (and congruency thereof), may result in significant medical error (and a lost opportunity to achieve patient-centred care).

Strengths of this study include the fact that we directly questioned patients and their family members in acute care settings on their views related to EOL treatment decisions. Patients were not questioned about various hypothetical scenarios, but were questioned about their actual current care. In addition, the multicentre nature of our study enhances the generalisability of our findings. A potential limitation of this study was the nature of some of the questionnaires used in our interviews with patients and their families. It may be that patients and families struggle to define their authentic values and preferences, or it may be a measurement problem that the scales used in this study were not adequately validated or do not adequately capture the complexities inherent in critical decision-making or a combination of both. There is minimal published work on the psychometric properties of the EOL values scale.15 Respondents rated the importance of each value in isolation of each other, with no attempt made to have respondents resolve discordant statements or rank all the values. There was considerable ‘ceiling effect’ that may have limited its ability to discriminate between preferences. Also, the tool we used to measure decisional conflict was an adaptation of a previously validated scale.13 We expanded the list of items on the scale to obtain information about decisional conflicts. These additions have face and content validity, but additional work was not carried out to understand the psychometric properties of this modified scale. Finally, our study population was largely English speaking and Caucasian. We are unsure how well our findings will generalise to other populations with more language, cultural and ethnic diversity.


Currently, there is considerable discordance between patients' stated values and preferences related to the use of life-sustaining treatments. Even when patients and family members have made a decision, there is considerable decisional conflict. Efforts to improve ACP and goals of care decisions for older, hospitalised patients are warranted. These should focus on the wide-spread implementation of values clarification tools and decision aids to better prepare patients and their families for ‘in the moment’ treatment decisions near or at the EOL.


The authors would like to acknowledge the contributions of the other members of the ACCEPT study team who are listed in a prior publication.10



  • Collaborators ACCEPT Study Team and the Canadian Researchers at the End of Life Network (CARENET).

  • Contributors DKH and AGD have had full access to all of the data in the study, and take responsibility for the integrity of the data and the accuracy of the data analysis. All authors contributed to the conception and design of the study and to the drafting of the article. AGD and XJ were responsible for data analysis. All authors contributed to the analysis and interpretation of data, revised the article for important intellectual content, and gave final approval of the version to be published. PD and JJY were also involved in data collection.

  • Funding This study was supported by funding from the Canadian Institutes of Health Research, The Michael Smith Health Services Research Foundation in British Columbia, Alberta Innovates and the AFP Innovation Fund in Ontario. The funders played no role in the design and conduct of the study; collection, management, analysis and interpretation of the data; and in the preparation, review or approval of the manuscript.

  • Competing interests None declared.

  • Ethics approval Research Ethics Boards of all participating institutions.

  • Provenance and peer review Not commissioned; externally peer reviewed.