Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2014 Jul 1.
Published in final edited form as: J Pain Symptom Manage. 2012 Sep 24;46(1):9–19. doi: 10.1016/j.jpainsymman.2012.07.002

Life-Sustaining-Treatment Preferences: Matches and Mismatches Between Patients’ Preferences and Clinicians’ Perceptions

Lois Downey 1, David H Au 1, J Randall Curtis 1, Ruth A Engelberg 1
PMCID: PMC3534846  NIHMSID: NIHMS397872  PMID: 23017611

Abstract

Context

Better clinician understanding of patients’ end-of-life-treatment preferences has the potential for reducing unwanted treatment, decreasing health care costs, and improving end-of-life care.

Objectives

To investigate patient preferences for life-sustaining therapies, clinicians’ accuracy in understanding those preferences, and predictors of patient preference and clinician error.

Methods

Observational study of 196 male veterans with chronic obstructive pulmonary disease (COPD) who participated in a randomized trial. Measures included patients’ preferences for mechanical ventilation (MV) and cardiopulmonary resuscitation (CPR) if needed in their current state of health, and outpatient clinicians’ beliefs about those preferences.

Results

Analyses were based on 54% of participants in the trial who had complete patient/clinician data on treatment preferences. Patients were more receptive to CPR than MV (76% vs. 61%; P<0.001). Preferences for both treatments were significantly associated with the importance patients assigned to avoiding life-sustaining therapies during the final week of life (MV: b=−0.11, P<0.001; CPR: b=−0.09, P=0.001). When responses were dichotomized (would/wouldn’t want treatment), physicians’ perceptions matched patient preferences in 75% of CPR cases and 61% of MV cases. Physician errors increased as patients preferred less aggressive treatment (MV: b=−0.28, P<0.001; CPR: b=−0.32, P<0.001).

Conclusion

Clinicians erred more often about patients’ wishes when patients did not want treatment than when they wanted it. Treatment decisions based on clinicians’ perceptions could result in costly and unwanted treatments. End-of-life care could benefit from increased clinician-patient discussion about end-of-life care, particularly if discussions included patient education about risks of treatment and allowed clinicians to form and maintain accurate impressions of patients’ preferences.

Keywords: treatment preferences, patient-clinician communication, chronic obstructive pulmonary disease, palliative care

Introduction

The disproportionate costs of medical care during the last year of life, particularly in the final month, have been widely noted (1–3). Almost two decades ago, researchers documented higher resource utilization and costs of care when physicians misunderstood their patients’ preferences to forgo life-sustaining therapies. They speculated that increased physician-patient communication, aimed at increasing both patients’ understanding of treatment options and physicians’ understanding of their patients’ preferences, might reduce health care expenditures (4). Researchers recently associated the occurrence of physician-patient discussions about end-of-life care with decreases in treatment costs, including costs associated with life-sustaining therapies (LST), and with improved quality of death (5). End-of-life care discussions are associated with less aggressive care and better patient outcomes at the end of life (6), and discussions about patients’ goals, values, beliefs, and treatment preferences improve adherence to patient wishes and patient/family satisfaction with care (7). End-of-life care discussions may serve both to educate patients about risks associated with some LSTs and thereby reduce their preference for these treatments, and to increase physician awareness of patients’ preferences to forgo treatment. In both ways, discussions have the potential for reducing unwanted treatment, decreasing health care costs, and improving quality of end-of-life care.

Importantly, studies also have documented that although many patients prefer end-of-life care focused on comfort rather than prolongation of life (8–14), a substantial minority prefer LST irrespective of the probability of adverse outcomes (12, 15, 16). Therefore, it is important that clinicians have an understanding of each patient’s preferences for end-of-life care.

Clinicians’ accuracy about patients’ treatment preferences is highest when a majority of patients have the same preference (17, 18), suggesting that judgments are based on what people in general want, rather than on individual patients. Clinicians’ perceptions of their patients’ preferences also may correspond with their own preferences for treatment (19, 20).

In this study, we used a sample of male veterans with moderate to severe chronic obstructive pulmonary disease (COPD) to examine patient preferences for two life-sustaining treatments – mechanical ventilation (MV) and cardiopulmonary resuscitation (CPR) – and their clinicians’ perceptions of those preferences. We focused on outpatient clinicians for several reasons: 1) their conversations with patients about end-of-life treatment issues are mandated quality-of-care markers (21), and their beliefs about patients’ treatment preferences may shape the approach to such discussions; 2) conversations between outpatient clinicians and patients are associated with the delivery of desired care, enhanced family outcomes, and reduced costs at the end of life (22, 23); and 3) outpatient clinicians’ perspectives may in some circumstances inform acute care clinicians and family members when patients become too ill to speak for themselves. Our goal was to identify factors associated with patient preferences and with clinician errors in perception.

Methods

Study Setting

The study is based on secondary analyses of data gathered for a trial testing an intervention to improve the quality of end-of-life discussions between patients with COPD and their primary care clinicians at two Veterans Affairs (VA) facilities in western Washington State. Details of the study are available elsewhere (24–26). In brief, the trial enrolled 92 clinicians and 376 patients with COPD between 2004 and 2007. The University of Washington Human Subjects Division and the VA Puget Sound Health Care System approved the study protocol, and all patients provided informed consent.

Each patient was asked to complete a baseline self-administered survey covering preferences for LSTs (16, 20, 27), history of discussions with the primary clinician about treatment preferences, self-assessed physical and mental health status, and other related topics. The patient identified his/her primary clinician on the survey, and the identified clinician was then asked to complete a baseline survey specific to the patient, including the clinician’s beliefs about the patient’s LST preferences and confidence that s/he understood the patient’s preferences. The measures used in the current study were collected before randomization, and we merged data from the treatment and control groups for analysis. We excluded women (n=11) from the current study because we believed this sample was too small to draw meaningful conclusions about outcomes of interest among women, or about possible associations between sex and the outcomes of interest. We included all male patients who provided information about their preferences regarding MV or CPR and whose clinician expressed a view about the patient’s preferences. If either the patient or clinician gave the response “don’t know,” the pair was excluded. Very few clinician surveys were excluded because of “don’t know” responses (13 clinician surveys [4%] for MV; 11 [3%] for CPR); most of the exclusions resulted from the clinician skipping the item (141 clinician surveys [39%] for both therapies). The resulting sample included 196 patients and their 68 clinicians.

Outcome Measures

The patient survey included the following question regarding MV: “If you were in your current health and unable to breathe on your own, would you want to be on a breathing machine for a few days? There would be no guarantee that you would be able to come off the breathing machine and be able to breathe on your own.” The parallel clinician item was, “In his current health, do you think [this patient] would want hospitalization in an ICU with short-term mechanical ventilation with an uncertain chance of being extubated alive?” Regarding CPR, the patient question was: “CPR would consist of electric shocks to the heart, pumping on the chest, help with breathing and heart medications given through the veins. Possible side effects of CPR include broken ribs and memory loss. In your current health, would you want CPR if your heart were to stop beating?” The clinician question was, “In his current health, do you think [this patient] would want CPR if he were to have a cardiac arrest?” For all four questions, the response options were 1 (definitely no), 2 (probably no), 3 (probably yes), and 4 (definitely yes).

In order to compare clinicians’ beliefs about their patients’ preferences with the patients’ actual preferences, we paired the clinician-patient responses and computed difference scores for each pair. Looking separately at MV and CPR preferences, we computed two measures of clinician error: 1) a directional error measure, clinician response minus patient response (−3 = maximum underestimation … 0=accurate assessment … +3=maximum overestimation); and 2) a non-directional error measure, the absolute value of the directional measure (0=accurate assessment … 3=maximum error). For example, if the clinician perceived that the patient would definitely want the treatment, and the patient indicated that he would definitely not want it, the score on the directional measure would be +3 (maximum overestimation); if the clinician perceived that the patient would definitely not want the treatment, but the patient indicated that he would definitely want it, the score would be −3 (maximum underestimation). In either of the example cases, the score on the non-directional measure would be 3. (We also computed an alternative dichotomous version of the patient and clinician responses, contrasting probably/definitely no with probably/definitely yes, and computed error scores based on this simplified version of the responses. However, results of the analyses with the alternative version were essentially the same as those obtained with the more nuanced version described above and are not reported in detail in this article.)

Predictors and Descriptive Measures

There were three primary predictors in our examination of clinicians’ prediction errors. Two were appropriate for use with both the directional and non-directional outcome measure: patients’ “no/yes” indication of whether a face-to-face discussion with the clinician about treatment preferences had ever occurred (27–31), and clinicians’ confidence in their knowledge of the patient’s general treatment preferences. The latter represents the clinician’s response to the question, “Do you feel you know the kinds of treatments your patient would want if he got too sick to speak for himself?” Response options were 1 (definitely no), 2 (probably no), 3 (probably yes), and 4 (definitely yes). The third primary predictor, appropriate only with the non-directional outcome, was the patient’s actual preference regarding treatment. This predictor was inappropriate for the directional outcome because its value imposes limitations on the possible values directional error can take (e.g., for patients whose response is “definitely no,” errors will always be in the positive direction; errors made for patients whose response is “definitely yes” will always be negative).

Variables from the patient’s baseline survey used in descriptive analyses and as possible predictors of outcomes of interest included age; racial/ethnic minority status (white non-Hispanic vs. minority); marital status (not currently married vs. currently married); general health status (a weighted pseudo-interval-scale score, 0–100, with 100 reflecting better health) and mental health status (a five-item short form Mental Health Inventory (MHI-5) score, 0–100, with 100 reflecting better health) (32); four St. George’s Respiratory Questionnaire scores (symptoms, activity, impact, and total; scale scores potentially ranging from 0=no disease to 100=maximum severity) (33), and the importance the patient assigned to avoiding ventilation or dialysis during the final week of life (0=least important to 10=most important) (11). From the clinician baseline survey, we used sex, racial/ethnic minority status, clinician type (physician vs. nurse practitioner, clinical nurse specialist, or physician assistant), and clinic affiliation (pulmonary care, general internal medicine, or other). Finally, to describe the patient sample, we report the number of discussions with the clinician about end-of-life treatment and the quality of those discussions (0=the very worst I could imagine to 10=the very best I could imagine).

Statistical Analysis

We tested associations between paired patient and clinician responses, using bivariate probit regression models with a weighted mean- and variance-adjusted least squares estimator, and with patients clustered under clinicians. We also used this modeling technique in testing patient-level factors associated with LST preference and with the non-directional measure of clinician error. For testing patient factors associated with the directional error score, we used linear regression with a restricted maximum likelihood estimator. To test clinician-level predictors, we substituted multi-level models, using probit or linear models as appropriate to the scale of the outcome being tested.

Tests for differences in central tendency between paired responses (e.g., the clinician vs. the patient response to a specified therapy, or the difference between the patient’s responses to the two therapies) were based on clustered Wilcoxon signed rank tests (34).

We used SPSS 17.0 (SPSS Inc., Chicago, IL) for descriptives, SAS 9.2 (SAS Institute Inc., Cary, NC) for clustered Wilcoxon tests, and Mplus 6.1 (Muthén & Muthén, Los Angeles, CA) for clustered and multilevel regression models. We report all P-values <0.05. However, in view of the large number of associations tested, we elected a priori to use the phrase “statistically significant” for associations having P-values <0.001.

Results

Characteristics of the Patient and Clinician Samples

Among the 196 patients and their 68 clinicians included in this study, 183 patients (68 clinicians) provided information about patients’ preferences for MV, and 188 patients (67 clinicians) provided information about preferences for CPR (Table 1). The 196 patients ranged from 39 to 91 years old, with 12% identifying with racial-ethnic minority groups. Their self-reported health status reflected wide variation.

Table 1.

Description of Samplesa

Patients
 Age, mean (SD) 68.6 (9.6)
 Racial/Ethnic Minority, n (%) 22 (12)
 Currently Married, n (%) 83 (42)
 St. George’s Severity Scores, mean (SD)b
  Symptoms 61.9 (20.6)
  Activity 65.0 (21.2)
  Impact 37.4 (16.3)
  Total 50.0 (15.9)
 Discussions with Clinician about End-of-life Treatment
  Had any discussions, n (%) 27 (15)
  Number of discussions, mean (SD) 0.5 (1.9)
  Quality of discussions, mean (SD)c 8.4 (1.6)
 General Health Status, mean (SD)d 43.9 (27.6)
 Mental Health Status, mean (SD)d 76.0 (19.5)
 Importance of Avoiding Ventilation/Dialysis in Last Week of Life, mean (SD)e 7.8 (3.2)
 Clinician’s Confidence in Knowledge of Patient’s General LST Preferences
  Confidence Rating, mean (SD)f 2.8 (0.8)
  Probably or Definitely Knows, n (%) 133 (71)
 Mechanical Ventilation with Current Health
  Patient’s Actual Preference
   Preference Rating, mean (SD)f 2.7 (1.2)
   Probably or Definitely Wants Treatment, n (%) 111 (61)
  Clinician’s Perception of Patient’s Preference
   Rating, mean (SD)f 2.9 (0.9)
   Probably or Definitely Wants Treatment, n (%) 129 (71)
 Cardiopulmonary Resuscitation with Current Health
  Patient’s Actual Preference
   Preference Rating, mean (SD)f 3.0 (1.1)
   Probably or Definitely Wants Treatment, n (%) 142 (76)
  Clinician’s Perception of Patient’s Preference
   Rating, mean (SD)f 3.0 (0.8)
   Probably or Definitely Wants Treatment, n (%) 151 (80)
Clinicians
 Female, n (%) 38 (56)
 Racial/Ethnic Minority, n (%) 17 (26)
 Physician, n (%) 47 (69)
Open in a new tab
a

The full sample included the paired group of 196 patients and their 68 clinicians who provided both a patient preference and clinician perception of that preference for either mechanical ventilation or cardiopulmonary resuscitation. Patient sample sizes for variables with missing data: patient minority status 182, St. George’s symptoms 182, activity 183, impact 183, total 181; any discussions 179; no. discussions 175; discussion quality 25; physical health status 186; mental health status 186; life support in last week of life 171; clinician’s confidence in knowledge of LST preferences 188; ventilation preferences 183; CPR preferences 188.

b

Potential range: 0 (no disease) to 100 (maximum severity).

c

Potential range: 0 (the very worst I could imagine) to 10 (the very best I could imagine).

d

Potential range: 0 to 100, with high scores reflecting better health.

e

Potential range: 0 (not important) to 10 (extremely important).

f

Potential range: 1 (definitely no) to 4 (definitely yes).

Over half of the clinicians were women, and almost 70% were physicians, with the remainder nurse practitioners, clinical nurse specialists, or physicians’ assistants. Most specialized in general internal (54%) or pulmonary medicine (31%), with a few from geriatrics or other primary care clinics.

A majority of both patients and clinicians reported that the patients would probably/definitely want MV or CPR if needed in their current state of health but, on average, patients assigned reasonably high importance to avoiding MV during the week before death. Although the vast majority of patients (almost 85%) reported never having had a discussion about end-of-life treatment issues with their clinician, more than 70% of the clinicians believed they probably/definitely knew the patient’s general LST preferences. Patients revealed substantial uncertainty regarding their preferences, with 43% in the “probably” (rather than “definitely”) categories for MV and 37% for CPR. Clinicians were even more likely to place their patients in the “probably” categories: 66% for MV and 67% for CPR.

Predictors of Patient Preference

Although patients were asked to indicate whether they would want MV or CPR in their current state of health, responses to these questions were not associated with the patients’ perceptions of their current health status. Those who rated their current health as reasonably poor were just as likely to prefer LST as were those who rated their health more favorably. In fact, the only strong predictor of patients’ preference for MV and CPR was the importance they assigned to avoiding MV and dialysis during the final week of life. With each increase of one point in the importance rating, patients’ scores decreased on the 4-point LST preference ratings by approximately 0.1 points (for MV, b = −0.11, P < 0.001; for CPR, b = −0.09, P = 0.001).

Clinicians’ Knowledge of Patient Preferences

Associations between patients’ LST preferences and their clinicians’ beliefs about those preferences (Fig. 1) were positively associated, although none achieved statistical significance using the strict P < 0.001 criterion. Errors in clinicians’ predictions were most commonly in the direction of clinicians’ over-attributing a preference for treatment, although this bias was not statistically significant.

Figure 1. Direction and amount of clinician error regarding patient preferences.

Figure 1

Open in a new tab

* Based on a clustered probit regression model (clinician perception regressed on patient preference), using a mean- and variance-adjusted least squares estimator

** Based on clustered Wilcoxon test

Predictors of Clinician Error

For each of the two types of LST, the amount of clinician error about patient preferences (i.e., without considering the direction of error) increased significantly as the patient’s reported preference for the treatment decreased (Table 2). Probit regression models for each therapy showed the predicted probability of clinician error was higher for patients who indicated that they probably/definitely did not want the treatment than for those who indicated they probably/definitely wanted it (Fig. 2).

Table 2.

Bivariate Associations Between Primary Predictors and Inaccurate Clinician Beliefsa

Outcome Predictor MV
CPR
b 95% CI R2 P b 95% CI R2 P
Non-directional Error
Extent to which patient prefers treatment (0–3) −0.28 −0.45, −0.12 0.10 <0.001 −0.32 −0.48, −0.15 0.11 <0.001
Any discussions of LST preferences (0–1)b 0.11 −0.31, +0.52 0.00 0.62 0.32 −0.16, +0.79 0.01 0.19
Clinician confidence about knowledge (0–3)c −0.11 −0.35, +0.14 0.01 0.40 −0.17 −0.38, +0.05 0.02 0.12
Directional Error
Any discussions of LST preferences (0–1)b 0.28 −0.24, +0.80 0.01 0.30 0.24 −0.28, +0.75 0.01 0.37
Clinician confidence about knowledge (0–3)c 0.21 −0.12, +0.53 0.01 0.21 −0.01 −0.23, +0.22 0.00 0.95
Open in a new tab
a

All estimates are based on single-predictor regression models. Models of non-directional error are based on probit regression, using a weighted mean- and variance-adjusted least squares estimator; models of directional error are based on linear regression, using a restricted maximum likelihood estimator. For patient-level predictors, patients were clustered under clinicians. For clinician-level predictors, the models were multi-level, with the predictor entered at the higher (clinician) level. Unless otherwise specified, the MV models were based on 183 patients and 68 clinicians; for CPR, there were 188 patients and 67 clinicians.

b

The model for MV was based on 168 patients and 67 clinicians; the CPR model on 174 patients and 66 clinicians with information about whether discussions occurred.

c

The model for MV was based on 175 patients and 67 clinicians; and for CPR on 180 patients and 66 clinicians, for which the clinician provided a response to the question, “Do you feel you know the kinds of treatments that your patient would want if he got too sick to speak for himself?” The question was not specific to MV or CPR, but addressed the clinician’s knowledge of the patient’s general treatment preferences.

Figure 2. Expected probability of clinician accuracy/errora by patient preference regarding treatment.

Figure 2

Open in a new tab

aSimplified error computation, in which both the patient’s preference and clinician’s perception of that preference were first dichotomized (probably or definitely would not desire the treatment vs. probably or definitely would desire it) before comparing the two values.

In contrast to the strong effect of patient preference for treatment on clinician accuracy, neither the occurrence of discussions about treatment preferences nor clinicians’ confidence that they were aware of their patients’ preferences showed an association with the amount or direction of error (Table 2). Only two tested predictors other than patients’ preferences produced P-values <0.05 in regression models predicting clinician error, and none met our more stringent (P < 0.001) test for statistical significance. Clinicians made somewhat fewer non-directional errors when the patient belonged to a racial-ethnic minority group (for MV, b = −0.42, P = 0.03; for CPR, b = −0.42, P = 0.04). In models of directional error, female clinicians were less likely than their male counterparts to over-perceive a preference for MV (b = −0.47, P = 0.04).

Preferences for MV vs. CPR

Among the 175 patients who indicated their preferences for both MV and CPR in their current state of health, 80 (46%) specified different preferences for the two therapies. Among these patients, there was a significant preference for CPR over MV, a clustered Wilcoxon test producing P<0.001. Although clinicians predicted a slight patient preference for CPR over MV, the difference was not statistically significant (P = 0.48 for the 4-point scoring).

Discussion

A majority of patients in this study indicated that, in their current health, they would want MV if they became unable to breathe on their own, or CPR if their heart stopped beating. However, their preferences for treatment were independent of their own assessments of their current health status, a pattern also noted in other research (16). In contrast, the importance patients attributed to avoiding MV and dialysis during the last week of life was strongly associated with rejection of MV and CPR in current health, suggesting that these seriously ill patients had attitudes toward LST that were independent of their immediate health status and were more likely to reflect beliefs and preferences that might remain stable even as their condition deteriorated. Many patients indicated some tentativeness in their preference, a pattern also reported in a study of chronically ill older adults (35), and consistent with the finding of a recent study, which reported frequent revisions in patient treatment preferences, even over relatively short time periods (36).

In a majority of cases, clinicians believed their patient would want MV or CPR if needed. However, clinicians suspected even more patient uncertainty than the patients reported themselves, putting about two-thirds of the patients in a category indicating a probable, but not definite, preference. Clinicians were more often correct than incorrect in their beliefs about whether their patients would want MV or CPR if needed in current health, based on a simplified yes/no choice. However, paired patient-clinician responses revealed many clinician errors in perception, with 39% in error about patients’ preferences for MV and 25% about preferences for CPR. Over 70% of the clinician surveys indicated that the clinicians believed they knew the patient’s general LST preferences. However, clinicians’ confidence in their knowledge was not significantly associated with accuracy.

Our study confirmed other reports that end-of-life treatment discussions occur infrequently: only 15% of the patients recalled having ever had a discussion of this sort with their primary clinician. The low occurrence may have been one reason why we found no association between patients’ reports of the occurrence of discussions and a reduction in clinicians’ misperceptions about patient preferences.

The only statistically significant predictor of clinician error was patient preference. Clinicians made significantly more errors as patients’ preference for treatment decreased. This pattern is consistent with a finding from SUPPORT: among patients who preferred a focus on extending life, only 2% felt their care ran counter to this approach; among those who preferred a palliative focus, 47% believed their care was inconsistent with their preferences (15). A recent exploration similarly found physicians significantly overestimating patients’ preferences for LST (37). These findings may reflect a general clinical default of providing LST unless there is an explicit refusal from the patient or family.

Patients in this study were significantly more accepting of CPR than of MV. A similar pattern was reported by patients with lung cancer or COPD in the SUPPORT study (10). Other studies have failed to replicate the finding, with some reporting roughly equal acceptance of the two therapies (9, 16), and at least one showing somewhat greater receptivity to short-term MV than to CPR (30). Three points are worth noting with regard to our study’s finding. First, researchers have reported that physicians sometimes use patients’ do-not-resuscitate orders as guides for decision making about other LSTs (38, 39). Our study suggests, as has other research (40, 41), that this may not always be a reliable guide, as patients may have differing preferences for specific LSTs. Second, there is a developing consensus that advance care planning should focus on identifying patients’ goals and values, rather than on documenting specific treatment preferences (42, 43). Our study provides insight that a patient may have differing preferences regarding differing therapies. Although, in some instances, an understanding of the patient’s overall goals and values may provide sufficient guidance for correctly predicting the preference for specific treatments, this may not always be the case. It is important that preferences for specific therapies be documented. Third, the fact that patients reported significantly greater acceptance of CPR than of MV suggests the need for education about the risks and benefits of different therapies during discussions about end-of-life treatment. In fact, survival after CPR is considerably lower than survival after MV for an acute exacerbation of COPD (44, 45). At least one study has shown that patients’ own estimates of their chances of surviving CPR are substantially higher than the estimates of their clinicians and that patients have very limited understanding of both CPR and MV (18). This may be, in part, a function of media depictions. Researchers studying portrayals of CPR on television, in which survival rates are in excess of those reported in the medical literature, have speculated that these portrayals may lead the public to an unrealistic impression of the chances of favorable CPR outcomes (46). It seems possible that patients in our study did not fully understand the details and implications of CPR and MV, and this may be part of the explanation for their higher receptivity to CPR.

We believe our finding of substantial misperception of patient preference by outpatient clinicians is important for several reasons. First, although primary clinicians are only rarely specifically asked to articulate patient preferences when patients lose decisional capacity, occasional instances occur. More importantly, however, primary clinicians’ misperceptions may influence the ways in which they present information about treatment options to patients and families; this may lead to decisions that are inconsistent with patients’ values and goals. In addition, we found clinician confidence in their knowledge of patients’ LST preferences to be unrelated to accuracy of perception; when such confidence is misplaced, clinicians may fail to discuss treatment options adequately with patients. Initiating discussions with patients can serve to correct perceptual errors that exist. The extent of misperception among clinicians in our study, along with the patients’ low recall of discussions about treatment preferences, suggests that communication about preferences, and perhaps also about the values and goals that should guide treatment decisions, was inadequate. Increasing the occurrence of discussions and providing explicit documentation of preferences in the patient’s medical record are critical to improvements in end-of-life care.

An important limitation of this study is its reliance on responses from a relatively small sample restricted to male U.S. veterans with COPD. The sample restrictions may limit generalizability to women, persons with other (or no) life-limiting conditions, those seeking treatment in other health care systems, or persons living in other geographic regions. A second limitation relates to ambiguities surrounding patients’ and clinicians’ interpretation of questions about patients’ wishes for life-sustaining treatment. We cannot be certain that respondents’ interpretations always matched those of the developers of the instrument or of the authors of this article, or that patients and clinicians interpreted their parallel questions in the same way.

Our findings suggest the continuing need for investigation of methods for increasing the occurrence of end-of-life-care discussions. Given the frequency of patient revisions in treatment preference, even an increase in the occurrence of one-time discussions will likely be insufficient; discussions must recur over the course of the clinician-patient association. Such discussions must include not only a focus on goals and values, as has been emphasized by experts on advance care planning, but also consideration of preferences for specific treatments. They also must include education of patients about benefits and, importantly, risks of these treatments; and they must incorporate methods whereby clinicians can form and maintain accurate impressions of their patients’ evolving treatment preferences. Finally, future development of surveys would benefit from cognitive interviewing of both patients and clinicians to ensure concordance between survey developers and respondents regarding the meaning of questions and response options. The quality of end-of-life care, in general, could benefit from further research investigating methods for improving accuracy on the part of both patients and clinicians.

Acknowledgments

We are grateful to Edmunds M. Udris for his efforts in organizing and preparing the data used in this study.

Footnotes

Disclosures

Primary funding was from the Department of Veterans Affairs grant #IIR-02-292. Additional funding was provided by the National Institute of Nursing Research and the National Cancer Institute through grant #R01 NR009987. The role of the agencies that supported this study was limited to financial support. They did not participate in study design/conduct or in data collection/management/analysis, nor did they have a role in preparation/review of the manuscript. The views expressed are those of the authors and do not necessarily represent the opinions of the Department of Veterans Affairs, the National Institute of Nursing Research, or the National Cancer Institute. The authors have no ties at present, nor have they had ties in the recent past, that would suggest potential conflicts of interest.

Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

References

  • 1.Hoover DR, Crystal S, Kumar R, Sambamoorthi U, Cantor JC. Medical expenditures during the last year of life: findings from the 1992–1996 Medicare current beneficiary survey. Health Serv Res. 2002;37(6):1625–1642. doi: 10.1111/1475-6773.01113. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 2.Hogan C, Lunney J, Gabel J, Lynn J. Medicare beneficiaries’ costs of care in the last year of life. Health Aff. 2001;20(4):188–195. doi: 10.1377/hlthaff.20.4.188. [DOI] [PubMed] [Google Scholar]
  • 3.The Dartmouth Institute for Health Policy & Clinical Practice. [Accessed July 5, 2011];The Dartmouth atlas of health care. Available from http://www.dartmouthatlas.org/
  • 4.Teno JM, Hakim RB, Knaus WA, et al. Preferences for cardiopulmonary resuscitation: physician-patient agreement and hospital resource use. J Gen Intern Med. 1995;10(4):179–186. doi: 10.1007/BF02600252. [DOI] [PubMed] [Google Scholar]
  • 5.Zhang B, Wright A, Huskamp H, et al. Health care costs in the last week of life: associations with end-of-life conversations. Arch Intern Med. 2009;169:480–488. doi: 10.1001/archinternmed.2008.587. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 6.Wright A, Zhang B, Ray A, et al. Associations between end-of-life discussions, patient mental health, medical care near death, and caregiver bereavement adjustment. JAMA. 2008;300:1665–1673. doi: 10.1001/jama.300.14.1665. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 7.Detering KM, Hancock AD, Reade MC, Silvester W. The impact of advance care planning on end of life care in elderly patients: randomised controlled trial. BMJ. 2010;340:c1345. doi: 10.1136/bmj.c1345. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 8.Barnato AE, Herndon MB, Anthony DL, et al. Are regional variations in end-of-life care intensity explained by patient preferences? A study of the U.S. Medicare population. Med Care. 2007;45:386–393. doi: 10.1097/01.mlr.0000255248.79308.41. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 9.Nahm E-S, Resnick B. End-of-life treatment preferences among older adults. Nurs Ethics. 2001;8(6):533–543. doi: 10.1177/096973300100800607. [DOI] [PubMed] [Google Scholar]
  • 10.Claessens MT, Lynn J, Zhong Z, et al. Dying with lung cancer or chronic obstructive pulmonary disease: insights from SUPPORT. J Am Geriatr Soc. 2000;48(5 Suppl):S146–S153. doi: 10.1111/j.1532-5415.2000.tb03124.x. [DOI] [PubMed] [Google Scholar]
  • 11.Downey L, Engelbert RA, Curtis JR, Lafferty WE, Patrick DL. Shared priorities for the end-of-life period. J Pain Symptom Manage. 2009;37(2):175–188. doi: 10.1016/j.jpainsymman.2008.02.012. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 12.Steinhauser KE, Christakis NA, Clipp EC, et al. Factors considered important at the end of life by patients, family, physicians, and other care providers. JAMA. 2000;284(19):2476–2482. doi: 10.1001/jama.284.19.2476. [DOI] [PubMed] [Google Scholar]
  • 13.Fried T, Bradley E, Towle V. Assessment of patient preferences: integrating treatments and outcomes. J Gerontol B Psychol Sci Soc Sci. 2002;57:S348–S354. doi: 10.1093/geronb/57.6.s348. [DOI] [PubMed] [Google Scholar]
  • 14.Nishimura A, Mueller P, Evenson L, et al. Patients who complete advance directives and what they prefer. Mayo Clin Proc. 2007;82:1480–1486. doi: 10.1016/S0025-6196(11)61091-4. [DOI] [PubMed] [Google Scholar]
  • 15.Teno J, Fisher E, Hamel M, et al. Decision-making and outcomes of prolonged ICU stays in seriously ill patients. J Am Geriatr Soc. 2000;48(5 Suppl):S70–S74. doi: 10.1111/j.1532-5415.2000.tb03144.x. [DOI] [PubMed] [Google Scholar]
  • 16.Stapleton RD, Nielsen EL, Engelberg RA, Patrick DL, Curtis JR. Association of depression and life-sustaining treatment preferences in patients with COPD. Chest. 2005;127:328–334. doi: 10.1378/chest.127.1.328. [DOI] [PubMed] [Google Scholar]
  • 17.Coppola KM, Ditto PH, Danks JH, Smucker WD. Accuracy of primary care and hospital-based physicians’ predictions of elderly outpatients’ treatment preferences with and without advance directives. Arch Intern Med. 2001;161:431–440. doi: 10.1001/archinte.161.3.431. [DOI] [PubMed] [Google Scholar]
  • 18.Fischer GS, Tulsky JA, Rose MR, Siminoff LA, Arnold RM. Patient knowledge and physician predictions of treatment preferences after discussion of advance directives. J Gen Intern Med. 1998;13:447–454. doi: 10.1046/j.1525-1497.1998.00133.x. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 19.O’Donnell H, Phillips RS, Wenger N, et al. Preferences for cardiopulmonary resuscitation among patients 80 years or older: the views of patients and their physicians. J Am Med Dir Assoc. 2003;4(3):139–144. doi: 10.1097/01.JAM.0000064464.85732.45. [DOI] [PubMed] [Google Scholar]
  • 20.Norris WM, Nielsen EL, Engelberg RA, Curtis JR. Treatment preferences for resuscitation and critical care among homeless persons. Chest. 2005;127:2180–2187. doi: 10.1378/chest.127.6.2180. [DOI] [PubMed] [Google Scholar]
  • 21.Lanken PN, Terry PB, DeLisser HM, et al. An official American Thoracic Society clinical policy statement: palliative care for patients with respiratory diseases and critical illnesses. Am J Respir Crit Care Med. 2008;177:912–927. doi: 10.1164/rccm.200605-587ST. [DOI] [PubMed] [Google Scholar]
  • 22.Mack JW, Weeks JC, Wright AA, Block SD, Prigerson HG. End-of-life discussions, goal attainment, and distress at the end of life: predictors and outcomes of receipt of care consistent with preferences. J Clin Oncol. 2010;28:1203–1208. doi: 10.1200/JCO.2009.25.4672. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 23.Mack JW, Paulk ME, Viswanath K, Prigerson HG. Racial disparities in the outcomes of communication on medical care received near death. Arch Intern Med. 2010;170:1533–1540. doi: 10.1001/archinternmed.2010.322. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 24.Au DH, Udris EM, Engelberg RA, et al. A randomized trial to improve communication about end-of-life care among patients with COPD. Chest. 2012;141:726–735. doi: 10.1378/chest.11-0362. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 25.Slatore CG, Cecere LM, Reinke LF, et al. Patient-clinician communication: associations with important health outcomes among veterans with COPD. Chest. 2010;138(3):628–634. doi: 10.1378/chest.09-2328. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 26.Reinke LF, Slatore CG, Udris EM, et al. The association of depression and preferences for life-sustaining treatments in veterans with chronic obstructive pulmonary disease. J Pain Symptom Manage. 2011;41(2):402–411. doi: 10.1016/j.jpainsymman.2010.05.012. [DOI] [PubMed] [Google Scholar]
  • 27.Curtis JR, Patrick DL, Caldwell E, Greenlee H, Collier AC. The quality of patient-doctor communication about end-of-life care: a study of patients with advanced AIDS and their primary care clinicians. AIDS. 1999;13:1123–1131. doi: 10.1097/00002030-199906180-00017. [DOI] [PubMed] [Google Scholar]
  • 28.Patrick DL, Pearlman RA, Starks HE, et al. Validation of preferences for life-sustaining treatment: implications for advance care planning. Ann Intern Med. 1997;127:509–517. doi: 10.7326/0003-4819-127-7-199710010-00002. [DOI] [PubMed] [Google Scholar]
  • 29.Curtis JR, Patrick DL, Caldwell ES, Collier AC. Why don’t patients and physicians talk about end-of-life care? Arch Intern Med. 2000;160:1690–1696. doi: 10.1001/archinte.160.11.1690. [DOI] [PubMed] [Google Scholar]
  • 30.Pearlman RA, Cain KC, Starks H, et al. Preferences for life-sustaining treatments in advance care planning and surrogate decision making. J Palliat Med. 2000;3:37–48. doi: 10.1089/jpm.2000.3.37. [DOI] [PubMed] [Google Scholar]
  • 31.Knauft E, Nielsen EL, Engelberg RA, Patrick DL, Curtis JR. Barriers and facilitators to end-of-life care communication for patients with COPD. Chest. 2005;127:2188–2196. doi: 10.1378/chest.127.6.2188. [DOI] [PubMed] [Google Scholar]
  • 32.Ware JE., Jr . SF-36 health survey: manual and interpretation guide. Boston: The Health Institute, New England Medical Center; 1993. [Google Scholar]
  • 33.Jones PW, Quirk FH, Baverstock CM, Littlejohns P. A self-complete measure of health status for chronic airflow limitation. The St. George’s Respiratory Questionnaire. Am Rev Respir Dis. 1992;145(6):1321–1327. doi: 10.1164/ajrccm/145.6.1321. [DOI] [PubMed] [Google Scholar]
  • 34.Rosner B, Glynn RJ, Lee M-LT. The Wilcoxon Signed Rank Test for paired comparisons of clustered data. Biometrics. 2006;62:185–192. doi: 10.1111/j.1541-0420.2005.00389.x. [DOI] [PubMed] [Google Scholar]
  • 35.Sudore R, Schillinger D, Knight S, Fried T. Uncertainty about advance care planning treatment preferences among diverse older adults. J Health Commun. 2010;15:159–171. doi: 10.1080/10810730.2010.499982. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 36.Janssen DJA, Spruit MA, Schols JMGA, et al. Predicting changes in preferences for life-sustaining treatment among patients with advanced chronic organ failure. Chest. 2012;141:1251–1259. doi: 10.1378/chest.11-1472. [DOI] [PubMed] [Google Scholar]
  • 37.Barnato AE, Mohan D, Downs J, et al. A randomized trial of the effect of patient race on physicians’ intensive care unit and life-sustaining treatment decisions for an acutely unstable elder with end-stage cancer. Crit Care Med. 2011;39(7):1663–1669. doi: 10.1097/CCM.0b013e3182186e98. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 38.Smith C, Bunch O’Neill L. Do not resuscitate does not mean do not treat: how palliative care and other modalities can help facilitate communication about goals of care in advanced illness. Mt Sinai J Med. 2008;75:460–465. doi: 10.1002/msj.20076. [DOI] [PubMed] [Google Scholar]
  • 39.Beach M, Morrison R. The effect of do-not-resuscitate orders on physician decision-making. J Am Geriatr Soc. 2002;50:2057–2061. doi: 10.1046/j.1532-5415.2002.50620.x. [DOI] [PubMed] [Google Scholar]
  • 40.Hickman SE, Tolle SW, Brummel-Smith K, Carley MM. Use of the physician orders for life-sustaining treatment program in Oregon nursing facilities: beyond resuscitation status. J Am Geriatr Soc. 2004;52:1424–1429. doi: 10.1111/j.1532-5415.2004.52402.x. [DOI] [PubMed] [Google Scholar]
  • 41.Fromme EK, Zive D, Schmidt TA, Olszewski E, Tolle SW. POLST registry do-not-resuscitate orders and other patient treatment preferences. JAMA. 2012;307:34–35. doi: 10.1001/jama.2011.1956. [DOI] [PubMed] [Google Scholar]
  • 42.Billings JA. The need for safeguards in advance care planning. J Gen Intern Med. 2012;27(5):595–600. doi: 10.1007/s11606-011-1976-2. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 43.Sudore RL, Fried TR. Redefining the “planning” in advance care planning: preparing for end-of-life decision making. Ann Intern Med. 2010;153:256–261. doi: 10.1059/0003-4819-153-4-201008170-00008. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 44.Ehlenbach WJ, Barnato AE, Curtis JR, et al. Epidemiologic study of in-hospital cardiopulmonary resuscitation in the elderly. N Engl J Med. 2009;361:22–31. doi: 10.1056/NEJMoa0810245. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 45.Connors AF, Jr, Dawson NV, Thomas C, et al. Outcomes following acute exacerbation of severe chronic obstructive lung disease. Am J Respir Crit Care Med. 1996;154:959–967. doi: 10.1164/ajrccm.154.4.8887592. [DOI] [PubMed] [Google Scholar]
  • 46.Diem S, Lantos J, Tulsky JA. Cardiopulmonary resuscitation on television: miracles and misinformation. N Engl J Med. 1996;334:1578–1582. doi: 10.1056/NEJM199606133342406. [DOI] [PubMed] [Google Scholar]

RESOURCES