Patient–physician communication about code status preferences

A randomized controlled trial


  • Wadih Rhondali MD,

    1. Department of Palliative Care and Rehabilitation Medicine, The University of Texas MD Anderson Cancer Center, Houston, Texas
    2. Department of Palliative Care, Centre Hospitalier de Lyon-Sud, Hospices Civils de Lyon, Lyon, France
    Search for more papers by this author
  • Pedro Perez-Cruz MD,

    1. Department of Palliative Care and Rehabilitation Medicine, The University of Texas MD Anderson Cancer Center, Houston, Texas
    2. Programa Medicina Paliativa, Departamento Medicina Interna, Facultad de Medicina, Pontificia Universidad Católica de Chile, Santiago, Chile
    Search for more papers by this author
  • David Hui MD,

    1. Department of Palliative Care and Rehabilitation Medicine, The University of Texas MD Anderson Cancer Center, Houston, Texas
    Search for more papers by this author
  • Gary B. Chisholm MS,

    1. Department of Biostatistics, The University of Texas MD Anderson Cancer Center, Houston, Texas
    Search for more papers by this author
  • Shalini Dalal MD,

    1. Department of Palliative Care and Rehabilitation Medicine, The University of Texas MD Anderson Cancer Center, Houston, Texas
    Search for more papers by this author
  • Walter Baile MD,

    1. Department of Behavioral Science, The University of Texas MD Anderson Cancer Center, Houston, Texas
    Search for more papers by this author
  • Eva Chittenden MD,

    1. Department of Palliative Care, Massachusetts General Hospital, Boston, Massachusetts
    Search for more papers by this author
  • Eduardo Bruera MD

    Corresponding author
    1. Department of Palliative Care and Rehabilitation Medicine, The University of Texas MD Anderson Cancer Center, Houston, Texas
    • Corresponding author: Eduardo Bruera, MD, Department of Palliative Care and Rehabilitation Medicine, Unit 1414, The University of Texas MD Anderson Cancer Center, 1515 Holcombe Boulevard, Houston, TX 77030; Fax: (713) 792-6092;

    Search for more papers by this author

  • This study was partially reported at the annual meeting of the American Society of Clinical Oncology, Chicago, IL, June 1-5, 2012.

  • See editorial on pages 1938-40, this issue.



Code status discussions are important in cancer care, but the best modality for such discussions has not been established. The objective of this study was to determine the impact of a physician ending a code status discussion with a question (autonomy approach) versus a recommendation (beneficence approach) on patients' do-not-resuscitate (DNR) preference.


Patients in a supportive care clinic watched 2 videos showing a physician–patient discussion regarding code status. Both videos were identical except for the ending: one ended with the physician asking for the patient's code status preference and the other with the physician recommending DNR. Patients were randomly assigned to watch the videos in different sequences. The main outcome was the proportion of patients choosing DNR for the video patient.


A total of 78 patients completed the study, and 74% chose DNR after the question video, whereas 73% chose DNR after the recommendation video. Median physician compassion score was very high and not different for both videos. All 30 of 30 patients who had chosen DNR for themselves and 30 of 48 patients who had not chosen DNR for themselves chose DNR for the video patient (100% versus 62%). Age (odds ratio = 1.1/year) and white ethnicity (odds ratio = 9.43) predicted DNR choice for the video patient.


Ending DNR discussions with a question or a recommendation did not impact DNR choice or perception of physician compassion. Therefore, both approaches are clinically appropriate. All patients who chose DNR for themselves and most patients who did not choose DNR for themselves chose DNR for the video patient. Age and race predicted DNR choice. Cancer 2013;119:2067–2073. © 2013 American Cancer Society.


Cardiopulmonary resuscitation (CPR) is part of the standard of care for patients who experience cardiac arrest during hospital admissions. The “do not resuscitate” (DNR) order legally documents that patients do not wish to pursue CPR in the event of a cardiopulmonary arrest. CPR has a low success rate in patients with advanced cancer.[1] CPR can also have negative consequences in this population, such as physical distress, loss of dignity and family suffering with complicated bereavement.[2] Studies have shown that the majority of patients who survive after CPR die within days to weeks in the intensive care unit, and few of them regain their previous functional status.[3-5] Because of these consequences, DNR status is generally considered appropriate for patients with advanced cancer.[6] However, the prevalence of DNR orders for these patients is only approximately 50%.[7, 8]

Given that a majority of cancer patients die with mental impairment, conducting code status discussions earlier in the illness is highly important in this population.[6, 9, 10] Discussions about code status are generally stressful and difficult for both patients and clinicians involved.[11] These discussions typically include a description of the interventions, their effectiveness, and the patient's preference.

Not all patients feel comfortable expressing a code status preference. Although autonomous decision-making is highly valued in the health care environment, studies have shown that the proportion of patients who prefer a shared decision-making style is larger than those who prefer an active or passive role.[12-14] In this context, it is crucial that physicians explore patients' communication preferences in order to guide the conversation accordingly. However, physicians are not always accurate at assessing a patient's decision-making style.[14, 15] In fact, agreement between patient decision-making preference and physician's perception of this preference occur in only 45% of the cases.[14]

Some progress has been made in understanding the factors that influence patient–physician communication in the context of advanced care planning. The content and the manner of the message delivered, environmental factors, and both patient and physician characteristics influence end-of-life communication.[13, 16-21] However, the impact of the physician's communication strategy in DNR discussions has not been studied in randomized controlled trials.

The aim of this study was to determine the impact of a physician's communication style, promoting patient autonomy versus promoting beneficence, on patient preferences regarding code status by exposing patients to 2 video scenarios. We hypothesized that patients who received a recommendation regarding code status would more frequently choose not to be resuscitated as compared to patients who did not receive a recommendation.



Patients who attended the Supportive Care Clinic at the University of Texas MD Anderson Cancer Center, Houston, Texas, between August and October 2011 were screened and subsequently asked to participate if they were deemed eligible for this study. Patients were included if they were 18 years or older and had a diagnosis of advanced cancer (defined as locally advanced, recurrent, or metastatic disease) and were referred to the Supportive Care Clinic. Non–English-speaking patients and those with impaired cognition were excluded. The Institutional Review Board at The University of Texas MD Anderson Cancer Center approved this study, and all patients gave written informed consent.


Discussions regarding DNR occur in a large variety of circumstances. The variation in the characteristics of the physician, environment (outpatient clinics, emergency departments, and intensive care units), and clinical situations could make the interpretation of the impact of communication strategies impossible. Therefore, we produced a standardized scenario with professional actors, reflecting an encounter in an outpatient setting. This approach has been used by our team before.[17, 22]

We produced 2 videos involving a middle-aged male Caucasian physician discussing code status with a female Caucasian patient who has advanced cancer and a life expectancy of weeks to months. The sex and race of the 2 actors were chosen to reflect typical consultations at a cancer center. Both videos included 5 minutes of a code status discussion, and were identical in every way except for the last sentence. The first video ended with the physician asking the patient for her code status preference (ie, autonomy-driven approach: the “question” scenario) and the second video ended with the physician recommending DNR for the patient (beneficence-driven approach: the “recommendation” scenario).

Both videos involved the same professional actors who were blinded to the objective of the study. The script reflected components of ideal physician communication styles identified from research literature, including breaking bad news in an empathic manner, eliciting patient preferences toward decision-making,[14] and addressing patient concerns and information needs.[23, 24] The video's design, content, and structure was reviewed and edited for appropriateness and accuracy by 2 medical oncologists, 2 palliative care physicians, and 1 psychiatrist specializing in patient–physician communication.

Randomization and Blinding

Prior to randomization, patients were stratified according to the decision-making preferences questionnaire (DMPQ). The DMPQ is a 5-item questionnaire that contains questions regarding decision-making preferences and groups patients into 3 categories: active, passive, or shared. It has been previously used in studies on cancer patients.[14, 25] The aim of the stratification was to obtain 2 homogeneous groups in terms of decision-making preferences (active/passive/shared). According to the DMPQ, patients were then randomized between 2 possible video sequences based on computer-generated random numbers (complete randomization): the first video ending with a question and the second video ending with a recommendation (sequence A), or the first video ending with a recommendation and the second video ending in a question (sequence B).

In previous studies, we observed that the sequence patients watch a video had impact on their perception of the physician's compassion. These studies suggest that patients generally prefer the doctor they see in the second video.[23, 24] To control this effect, we used a crossover design to assess the effect of sequence order on ratings of compassion and overall preference.

Both the research assistant in charge of conducting the assessments and the patients were blinded to the allocation sequence throughout the study. Patients were blinded to the hypotheses of the study.


We collected patient data such as age, sex, ethnicity, religious affiliation, marital status, highest educational level, cancer diagnosis, and date of cancer initial diagnosis. Symptom burden was documented using validated Edmonton Symptom Assessment Scale (ESAS).[26-29] We also documented strength of religious faith using the Abbreviated Santa Clara Strength of Religious Faith Questionnaire,[30, 31] because religious faith has been reported to influence patient preferences regarding end-of-life.[32, 33] This tool has been validated for cancer patients.[34]

After the first and second video, all patients were asked for code status preference if they were the patient they just watched, their overall preference for communication style of the first physician on a 0-to-10 scale, and the physician compassion assessed by a validated 5-item tool consisting of 0-to-10 numerical rating scales (with 0 for warm and 10 for cold; likewise for pleasant/unpleasant, compassionate/distant, sensitive/insensitive, and caring/uncaring).[17, 35]

After these assessments, patients were asked if they have had a previous code status discussion with their doctor and if they have made a decision regarding CPR.

Statistical Considerations

Patient characteristics and baseline measurements were analyzed with simple parametric or nonparametric statistical methods as appropriate for each variable.

The primary objective of the study was to evaluate the difference in DNR choice of patients exposed to one of the 2 videos (after the first video). This primary objective was analyzed by contingency table analysis. A 2-sided chi-square test of equal proportions was performed. We estimated that with 78 patients (39 patients per arm), the study would have 80% power to detect an odds ratio of 4.0 or larger (50% chose DNR after watching the first video of the sequence A versus 80% chose DNR after watching the first video of the sequence B) when alpha = 0.05. Total accrual for the study will be 78 patients.

We analyzed the binary DNR preference variable using crossover methodology proposed by Kenward and Jones.[36] Briefly, a chi-square test was performed on a 2 × 2 contingency table that includes the patients who had a shift response after watching both videos. This chi-square test tested the intervention effect by comparing the response shifts between the 2 groups.

We analyzed compassion (and other secondary variables) using standard 2-stage crossover methodology.[37] All tests were 2-sided, and P values less than .05 were considered statistically significant. Analyses were performed using SAS version 9.3 (Cary, NC).


A total of 97 patients were approached for the study, 80 were finally included, and 78 patients completed the study (Fig. 1). Baseline demographics, symptom intensity, anxiety and depression, and religiosity were similar in both study arms (Table 1). The study population preferred a shared decision-making style over an active or passive decision-making style and the distribution was similar in both groups. Thirty patients (38%) stated that their code status was DNR and 8 patients (10%) said that their code status was CPR, and 40 patients (52%) from our study population had not made a decision regarding their own code status at the time of the study. Only 3 (4%) patients in our study had their code status explicitly written in their chart.

Figure 1.

Consort diagram is shown of the study process and participants' flow.

Table 1. Baseline Patient Characteristics
Patient CharacteristicsSequence A: (Q–R) N = 45Sequence B: (R–Q) N = 35P
  1. a

    t test.

  2. b

    chi-square test.

  3. c

    Fisher exact test.

  4. d

    Wilcoxon rank-sum test (2-sided).

  5. Abbreviations: CPR, cardiopulmonary resuscitation; DNR, do not resuscitate; ESAS, Edmonton Symptom Assessment System; IQR, interquartile range (Q1-Q3); Q, video ending with a question; R, video ending with a recommendation; SC, Santa Clara Religious Faith Questionnaire; SD, standard deviation.

Mean age (SD)53 (15)52 (13)0.81a
Female (%)25 (56)21 (60)0.69b
Married (%)28 (62)19 (54)0.47b
Highest education level (%)19 (42)16 (46)0.94b
High school or below21 (47)15 (43) 
Any college education5 (11)4 (11) 
Advanced degree   
Race (%)32 (71)27 (77)0.47c
White8 (18)7 (20) 
Black5 (11)1 (3) 
Religion (%)6 (13)9 (26)0.48b
Baptist8 (18)5 (14) 
Catholic9 (20)8 (23) 
Christian/Protestant22 (49)13 (37) 
Cancer diagnosis (%)10 (22)1 (3)0.05c
Lung9 (20)9 (26) 
Gastrointestinal8 (18)3 (9) 
Genitourinary3 (7)6 (17) 
Breast15 (33)16 (45) 
Time from cancer diagnosis in months, median (IQR)26 (13, 54)22 (10, 63)0.83d
ESAS, median (IQR)3 (2, 6)3 (1, 6)0.95d
Pain4 (2, 5)5 (3, 7)0.10d
Fatigue0 (0, 3)1 (0, 3)0.36d
Nausea2 (0, 4)2 (0, 4)0.51d
Depression2 (0, 4)2 (0, 4)0.88d
Anxiety2 (0, 5)3 (2, 4)0.06d
Drowsiness2 (0, 3)3 (1, 5)0.03d
Shortness of breath3 (1, 6)4 (1, 5)0.57d
Lack of appetite4 (2, 5)4 (3, 5)0.35d
Sleep4 (3, 5)4 (2, 5)0.79d
SC (religious questionnaire), median (IQR)18 (14, 20)16 (13, 20)0.40d
Decision-making style13 (29)9 (23)0.95b
Active27 (60)22 (66) 
Shared5 (11)4 (11) 
Do you have a code status? (n = 78)20 (44)20 (61)0.08c
No22 (49)8 (24) 
Yes, DNR3 (7)5 (15) 
Yes, CPR   

Fifty-eight patients chose DNR after the first video. There was no difference in the proportion of patients who chose DNR for the patient according to the type of video (Question video versus Recommendation video) after the first video in both sequence (34 of 58 [59%] versus 24 of 58 [41%], respectively; P = .49).

We did not find any significant association between DNR choice for the patient in the first video and symptom burden (ESAS; pain, P = .14; fatigue, P = .34; nausea, P = .82; depression, P = .61; anxiety, P = .70; drowsiness, P = .72; shortness of breath, P = .16; lack of appetite, P = .83; sleep, P = .34; well-being, P = .47) as well as with religious faith (P = .70) and decision-making preference style (P = .34).

The proportion of patients who chose DNR for the video patient was not influenced by the period the videos were watched (Period 1 versus Period 2: 56 (72%) versus 59 (76%); P = .58). For the crossover analysis, Table 2 describes the combinations of responses each patient had after watching both videos. The 2 groups were not significantly different in the effect the videos had on patient response shift regarding DNR preference (Table 2).

Table 2. Number of Patients Who Chose the 4 Possible Sequences of Responses After Watching Both Videos
  1. a

    Two-tailed Fisher exact test (P = 1).

  2. Abbreviations: DNR, do not resuscitate; Q, question video; R, recommendation video.

1 (Sequence Q-R)101a0a3445
2 (Sequence R-Q)83a1a2133

After watching the first video, the median patient rating for overall preference for communication (on a scale of 0 to 10) was 9 (interquartile range: 8, 10) and for compassion score (on a scale of 0 to 50) was 6 (interquartile range: 0, 13). There were no differences among the 2 groups (P = .93 and P = .53 for overall preference for communication and compassion score, respectively). There were no significant differences in the overall physician impression score according to the type of video watched or to the period the videos were watched (P = .73 and P = .86, respectively). There was no video or period effect for physicians' compassion score either (P = .45 and P = .47, respectively). Compassion score was significantly associated with DNR choice after the first video (P = .04), with patients who did not choose DNR giving a worse score than patients who chose DNR.

In order to identify predictor variables of choosing DNR for the video patients, we performed a subgroup analysis. Sixty patients (77%) chose, for at least one of the videos, a DNR order for the video patient (Table 2). A total of 30 of 30 patients who chose DNR for themselves chose DNR for the video patient, whereas 30 of 48 patients who did not choose DNR for themselves, chose DNR for the video patient (100% versus 62%, P < .0001) (Table 3). Patients who chose DNR for the video patient were older, more likely to be married, and were predominantly white (Table 4). In multivariate analysis that included age, race, sex, marital status, education, and religion, both age (odds ratio = 1.016 per year, P = .01) and race (odds ratio = 9.43, P = .004) were independent predictors.

Table 3. Comparison of Personal Code Status and Code Status Chosen for the Video Patient
 Code Status Chosen for the Video Patient  
Personal Code StatusNot DNR n (%)DNRa n (%)Total n (%)P
  1. a

    Patients were defined to have chosen “do not resuscitate” (DNR) for the video patient if they chose it either for patient in Video 1 or Video 2.

  2. b

    Chi-square. Value in bold denotes statistical significance.

Not DNR18 (37)30 (63)48 (62)<.001b
DNR0 (0)30 (100)30 (38)
Total18 (23)60 (77)78 (100) 
Table 4. Subgroup Analysis
 DNR Preference for Video PatientDNR Preference for Patients Themselves
 DNR (N = 60)No DNR (N = 18)PDNR (N = 30)No DNR (N = 48)P
  1. a

    t test.

  2. b

    Chi-square test.

  3. c

    Fisher exact test.

  4. Values in bold denote statistical significance. Abbreviations: DNR, do not resuscitate; SD, standard deviation.

Age, mean (SD)55 (14)45 (12).01a57 (12)50 (15).04a
Female, n (%) N = 4532 (71)13 (29).15b12 (27)33 (73).01b
Married, n (%) N = 4539 (87)6 (13).02b21 (47)24 (53).08b
RaceWhite, n (%) N = 5749 (86)8 (14).001c26 (46)31 (54).002c
Black, n (%) N = 156 (40)9 (60)0 (0)15 (100)
Other, n (%) N = 65 (83)1 (17)4 (67)2 (33)


Our study shows that an autonomy-driven versus a beneficence-driven approach to a code status discussion had no impact on patients' DNR preference or overall physician impression. These findings suggest that ending a DNR discussion with a question or a recommendation are both appropriate in the clinical setting. One possible explanation for this negative finding could be that the information provided by the physician in the video was more important than the way the physician ended the code status discussion. This is supported by the high preference for DNR for the video patient overall, even among those patients who did not choose DNR for themselves. Although the described scenario was one in which DNR was considered appropriate, almost 25% of the patients did not choose DNR with both modalities of communication. More research is needed to better understand factors that influence DNR choices among palliative care outpatients.

A second hypothesis to explain this negative result could be that the compassion and care that a physician provides could be more important than the way a code status discussion ends. Patients perceived the physician as extremely compassionate after watching both videos, suggesting that both approaches (autonomy-driven versus beneficence-driven) seemed to contain qualities highly valued by patients. It is important to highlight that the physicians' scores were higher in this study than what we have reported in other studies using the same tool (mean score [SD]: 41.6 [8.3] versus 29.5 [12.7]), as well as what was reported by the original authors who used this scale.[17, 35]

Finally, this negative result could be explained by the fact that the 2 videos were not different enough to show a difference in a patient's perception of the physician communication style and that changing the end of the conversation was not a sufficient way to evaluate these 2 communication strategies (patient autonomy versus beneficence). Future research should probably use drastically different videos to be sure that we assess different styles of communication for code status discussion.

A very interesting finding in our study is that all patients who had a DNR order for themselves decided they would choose DNR for the video patient. On the other hand, 62% of the patients who did not have a DNR order said they would choose a DNR order for the video patient. We hypothesize that the group of patients with advanced cancer who have not chosen a DNR order may have not discussed this issue with their primary doctor, but they are at a stage where they foresee that this alternative is a real and valid option in certain cases. Health care providers who take care of these patients may be missing invaluable opportunities to address this important issue.

In previous studies, we found that patients scored physician's compassion significantly higher for the physician observed in the second video.[17, 22] In the current study, the order in which the videos were shown did not modify the patient's preference. This finding could also be explained by the masking effect of a highly compassionate physician in the video.

When patients were compared according to whether they chose DNR for the video patient, being older, being married, and having white ethnicity were predictors of choosing DNR. In the multivariate analysis, both age and race were independent predictors of choosing DNR for the video patient. These results are consistent with previous literature by our group and others.[6, 38, 39] Several studies have reported increased rate of DNR documentation with age.[40-43] Other reports have consistently shown racial difference in DNR preferences. For example, a recent review described that white race is a predictor of having a DNR order.[40] Likewise, a recent study showed that black patients are less likely to have a code status documented in their medical record.[41]

We also found that most patients (51%) had not made a decision regarding their own code status and that only 4% of the patients had a code status noted in the chart. Our results are consistent with other publications that report only a small proportion of advanced cancer patients have a DNR order in their medical record despite evidence that shows advanced care planning and code status discussions is an important part of comprehensive care.[41, 44, 45] These results suggest that the remaining 96% with no documentation regarding their code status were exposed to full resuscitative measures, regardless of their preference.

Prior conversation with the oncologist regarding code status is a potential important covariate for this study. We documented patient prior conversation with their oncologist regarding code status; however, only 3 patients reported a prior discussion, and this low frequency did not allow us to use this variable in our multivariate analysis.

Regarding code status discussions, our results suggest that initiating this topic by a question or recommendation will not play an important role on patient final choice, but discussions should take place more frequently and they should be systematically documented. In our results, a large proportion of patients (51%) did not make any choice regarding their own code status, suggesting that it might require multiple conversations to allow patients to understand the goals and risks regarding resuscitation in the context of advanced cancer and to make their own choice.

Code status discussions should happen early during the course of illness for every patient, and one of the ways to improve the rate of patient-documented preferences for code status will be to refer the patient to supportive care teams and to increase educational procedures for both the public and physicians. Future research should study the impact of repeated conversations to see if patient DNR preference changes over time, particularly among outpatients who did not know about or who refuse DNR.

The main limitation of this study was that we tested the impact of a standardized video on patient preferences for code status. In clinical practice, physicians are likely to vary in terms of the clarity of the provided information as well as for the empathy of their communication style, resulting in differences in patients' preferences.[46] This might limit generalizability of our results to usual code status discussion.


Exposing outpatient cancer patients to 2 different videotaped patient–physician discussions regarding code status, one ending a DNR discussion with a question and the other one with a recommendation did not significantly affect patients' DNR decisions or perception of physician compassion, and therefore both approaches are clinically appropriate. Of all patients, 77% chose DNR for the video patient, even though most of them had not chosen DNR for themselves. A DNR order for the video patient was predicted by DNR decision for the patient themselves, older age, and being of white race. More research is needed to develop evidence-based communication strategies.


No specific funding was disclosed.


Dr. Bruera is supported in part by National Institutes of Health grants R01NR010162-01A1, R01CA1222292.01, and R01CA12 4481-01 and in part by the University of Texas MD Anderson Cancer Center support grant CA 016672.