Secondary Logo

Journal Logo

End-of-Life Care

A Controlled Trial of a Short Course to Improve Residents' Communication With Patients at the End of Life

Alexander, Stewart C. PhD; Keitz, Sheri A. MD, PhD; Sloane, Richard MPH; Tulsky, James A. MD

Author Information
doi: 10.1097/
  • Free


High-quality end-of-life care requires physicians who communicate effectively. Excellent communication skills are associated with the following improvements for patients: more successful pain control, improved emotional and psychological well-being, increased adherence to treatment recommendations, and enhanced satisfaction.1,2 Conversely, poor physician communication skills have been associated with patients' distress, confusion about medical treatment, and uncertainty about the medical information presented.3 The quality of communication also influences physicians by affecting their stress levels, job satisfaction, and emotional burnout.4 Poor communication has even been associated with increased rates of malpractice claims.5 A consensus exists among major health care organizations that medical education must prioritize communication training, particularly with regard to end-of-life care.6

Although effective communication skills have been identified as a key component in the delivery of excellent palliative care, many physicians still do not receive adequate training in tasks such as breaking bad news, addressing patients' emotional concerns, and discussing patients' preferences for life-sustaining treatment.6,7 In a national survey of 1,455 medical students and 296 third-year residents, respondents felt unprepared to address psychosocial, emotional, and spiritual issues with their patients.7 In addition, medical faculty and third-year residents from 62 medical schools felt unprepared to teach many of these key communication skills related to end-of-life care.7

To our knowledge, only a handful of controlled studies have objectively looked at the effect of educational interventions on specific end-of-life communication by analyzing audio-recorded conversations for specific communication behaviors used by health professionals.8–10 Razavi and colleagues9,10 showed that after an intervention, French and Belgian oncology nurses improved both the content and delivery of their interviews. Fallowfield et al.8 studied British oncologists attending a three-day intensive residential communication skills course and found significant improvement in their question-asking skills and their ability to respond to patient emotions.8

This research leaves open several questions. Will such training programs be successful with American health care providers? Are resident physicians as receptive to training as attendings? And, finally, can this training be offered in a setting that is less time intensive than a multiple-day residential retreat? We conducted this study to evaluate the effectiveness of a relatively brief end-of-life communication skills training program for internal medicine residents.



We invited 56 medical residents participating in the VA ambulatory care rotation at the Duke University Medical Center from 1999 to 2001 to attend an intensive palliative care educational retreat called PREPARE (Program of Resident Education to Promote Awareness and Respect at the End of life). Of these 56 residents, 37 comprised the intervention group. Nineteen residents from the same program who rotated at the VA and had not attended the retreat were eligible to be included in the control group.


The two-day retreat was composed of a 16-hour curriculum that focused on skills necessary for providing excellent end-of-life care. It included three components: control of pain and symptom management (six hours); communication skills (five hours); and sessions designed to promote residents' understanding of the experience of patients and families, enhance their personal awareness, and inform them about ethical issues (five hours). The communication skills work was all taught by the same instructor (JAT) and included one session each (90 minutes) on “delivering bad news” and “eliciting patient preferences for treatment/discussing DNR.” Each topic was introduced via small-group lecture/discussion and supported with a variety of audio-visual materials, including relevant clips from popular television programs, recordings of actual physician–patient encounters, and scripted “trigger tapes.” Learners then practiced the specific skills through role-play. These two topics were followed by a two-hour communication skills practice session during which learners were encouraged to bring in challenging cases from their own experience and to recreate the scenarios in further supervised role-play. For this study, we only analyzed the communication component of the intervention.

Evaluation measures

We evaluated residents' communication skills using audio-recorded encounters with standardized patients before and after the intervention. In these encounters, the physicians completed two tasks: delivering bad news and discussing patients' preferences for life-sustaining treatments. We developed four cases for each task, and carefully trained actors to portray these roles in a standardized manner. The residents in the intervention group interviewed four standardized patients before the intervention (two bad-news cases and two patient-preferences cases) and four standardized patients afterwards (again, two of each). The control physicians completed only one evaluation at one point in time (two bad-news cases and two patient-preferences cases). Standardized patient cases were assigned to all residents in random order, and residents never saw the same patient actor twice.

Two communication evaluation tools (Bad-News Conversations and Patient Preferences) were derived from empirical literature and were based on the “standards of practice” for palliative care.11–14 Both evaluation tools incorporated specific content and interaction physician communication skills.

Audio recordings of the bad-news conversations were evaluated in three categories: delivering bad news, responding to emotional cues, and general communication skills. Delivering bad news consisted of five specific skills: (1) issuing a “warning shot” before delivering the bad news; (2) delivering the news early in the conversation; (3) using clear and explicit language; (4) remaining silent after presenting the news; and (5) asking for the patient's emotional reaction to hearing the news. Responding to emotional cues consisted of four skills: (1) naming the patient's emotion; (2) validating the patient's emotion; (3) expressing understanding for the patient's emotion; and (4) exploring the patient's emotion. General communication consisted of six skills: (1) establishing initial rapport; (2) assessing patient's social support; (3) making a nonabandonment statement; (4) eliciting additional patient questions; (5) checking for immediate patient safety; and (6) summarizing a follow-up plan.

Encounters about patient preferences for life-sustaining treatments were evaluated for 14 skills in three categories: informing about advance directives, eliciting patient preferences, and general communication. Informing about advance directives consisted of five skills: (1) describing the purpose of an advance directive; (2) explaining why the discussion is being held now; (3) reassuring that death is not imminent; (4) informing patients that they may change their mind about what treatment they select; and (5) recommending that patients name a surrogate decisionmaker. Eliciting patient preferences consisted of four skills: (1) asking patients about prior experiences with end-of-life decision making; (2) asking about specific preferences; (3) presenting clinical scenarios; and (4) providing quantitative estimates of probability. General communication consisted of five skills: (1) assessing the patient's understanding of her or his illness; (2) addressing emotional content; (3) exploring emotional reactions; (4) checking understanding of information; and (5) providing an opportunity to ask questions.


Two research assistants (coders), who were blinded to physician treatment group, evaluated the encounters. Thirty-seven percent of the bad-news conversations and 40% of the patient-preferences conversations were coded by both raters and a kappa statistic was calculated for each code to obtain interrater reliability of the coding system. Disagreements between the coders were discussed and resolved, and final consensus was used for subsequent analysis. Most discrepancies were due simply to oversight by one or both coders. In rare cases when discrepancies remained, they were resolved by the entire research team. Bad-news codes demonstrated moderate to excellent reliability (κ = 0.49–0.86), and all patient-preferences codes demonstrated excellent reliability (κ > 0.80).

We assigned one point for each communication skill participants used during the encounter, with the exception of two skills: issuing a warning shot before delivering the bad news and remaining silent after presenting the news. These were the two skills that had received particular attention by the instructors when teaching bad-news delivery and practiced by all learners. Therefore, issuing a warning shot before delivering the bad news received a weighted score of 2 when performed, and remaining silent after presenting the news was given a score of 0, 1, or 2, depending on the length of the silence maintained by the resident following the bad-news delivery (with longer silences receiving higher scores).

We computed summary scores for each of the two tasks by adding all the averaged skills scores, creating a scale of 0–17 for bad-news conversations and 0–14 for patient-preferences conversations. In addition, we computed summary scores for each of the above listed subcategories. We computed final skills scores by combining the scores for bad-news and for patient-preferences conversations.

During analysis, we first conducted a two-sample t test of the intervention group's pre-retreat scores and the control group scores to assess for baseline differences between groups. We then conducted another two-sample t test to establish whether a difference existed between the intervention group's post-retreat scores and the control group scores. This was followed by a series of one-sample t tests in the intervention group to look at changes in the pre- and postintervention overall summary scores and subcategory scores for each scenario. An overall “change score” was computed for each participant by evaluating the difference between the postintervention summary score and the preintervention summary score for each of the subcategories.


All 56 residents participated in the study, for a 100% response rate. Nineteen were in the control group and 37 were in the intervention group. The residents included 26 interns (46%), 27 junior residents (48%), and three senior residents (5%) (Table 1). The group was composed of 27 women (48%) and 29 men (52%), and of the group, 60% were white, 7% were Latino, 4% were African American, and 29% were Asian. There were no differences between the control and intervention groups with regard to prior formal training in end-of-life care (P = .11); prior experience with delivering bad news (P = .25); number of hospice patients followed as primary caregiver (P = .58); number of patients for whom the physicians were present at the last hour of their death (P = .63); previous experience with the death of a family member (P = .49) or friend (P = .72); or prior life-threatening illness (P = .35).

Table 1
Table 1:
Baseline Characteristics of 19 Control-Group and 37 Intervention-Group Residents in a Study of End-of-Life Communication Skills, Duke University Medical Center, 1999–2001

Bad news

Respondents completed two standardized-patient scenarios; the first was delivering bad news. We found no difference in overall summary scores at baseline between the intervention and control groups (7.81 vs. 8.37; P = .25; highest possible score = 17). However, after the retreat, the intervention group summary scores were significantly higher than the control group (9.58 vs. 8.37; P = .04). In addition, the within-group testing showed an improvement of summary scores after the retreat in the intervention group (7.81 vs. 9.58; P = .001). In addition to these summary score findings, improvements were also noted in the subcategories of delivering bad news (2.45 vs. 3.73; P < .0001; highest possible score = 7); and responding to emotional cues (1.31 vs. 1.91; P = .003; highest possible score = 4) (Table 2). However, no significant difference was found for using general communication skills (4.05 vs. 3.94; P = .63; highest possible score = 6).

Table 2
Table 2:
Subcategory Scores for Bad News and Patient Preferences Encounters for 37 Intervention-Group Residents in a Study of End-of-Life Communication Skills, Duke University Medical Center, 1999–2001

Patient preferences

The second standardized-patient scenario involved eliciting patient preferences for life-sustaining care. We found no significant difference at baseline between the intervention and control groups' overall summary scores for patient-preferences conversations (5.90 vs. 5.84; P = .60; highest possible score = 14). There was also no significant difference between postintervention and control groups (6.31 vs. 5.84; P = .55), as well as between pre- and postintervention overall scores for this case (5.90 vs. 6.31; P = .83). In the subcategory analysis, a significant difference was noted in the subcategory of eliciting patient preferences (1.10 vs. 1.83; P < .0001; highest possible score = 4), whereas none was found in the other two subcategories of informing about advance directives (2.62 vs. 2.56; P = .83) and using general communication skills (2.18 vs. 1.94; P = .37; highest possible score = 6 for each subcategory) (Table 2).


This study demonstrates that a two-day, intensive palliative care retreat can improve communication skills of U.S. medical residents. We observed overall improvement in residents' delivery of bad news, as well as specific improvements in the way physicians delivered bad news and responded to patients' emotional cues. Although there was no overall improvement with regard to discussions about patient preferences for treatment, we noted a significant improvement in specific skills used by physicians in these conversations.

It is not surprising that the greatest improvement was found in the ability of medical residents to deliver bad news. This is a communication skill that is composed of very discrete components that are relatively easy to learn. The SPIKES protocol is an example of how these skills can be broken down into teachable segments.11 In contrast, discussions with patients about their preferences for life-sustaining treatments and advance care planning are more fluid, and thus less easily taught as a series of concrete communication tasks. These conversations are dependent on the facile use of general communication skills intermixed with techniques specific to end-of-life communication.14 Confirming this hypothesis was our observation that within the patient preference conversations, residents did improve in the subcategory of “eliciting patient preferences” and specifically in areas such as using quantitative estimates of probability. These were skills that were clearly enumerated in the training and that, apparently, resonated with the learners.

General communication skills did not improve in either of the standardized-patient cases. At least three potential explanations exist. First, although some of these skills were covered in the teaching sessions, they were not as much the focus of learning as were the higher-order skills specifically related to palliative care. Second, the instruction time was quite limited compared with other successful communication teaching interventions.8,15,16 These most “basic” skills require quite a bit of practice and review, for which insufficient time was available during our intervention. Finally, most residents have already been exposed to some general communication training during medical school, which may have had a greater effect on their performance than what was provided in our short course.

This study had several limitations. First, although the trial was controlled, the sample was not randomly assigned to intervention and control groups. Residency program scheduling restrictions did not allow this to occur. Nevertheless, there was little reason to believe that a significant difference existed among the residents in the two groups. We did not find a significant difference in baseline communication skills between the control and the intervention physicians. Overall, the skills were lower than we expected, and barring an unknown secular trend, improvements in performance were likely due to the normal and usual skills training. In addition, all of the available residents participated in the study, which eliminated a refusal bias. Additional limitations were that the sample was relatively small, the study was limited to only one site, and the communication training employed only one instructor. This intervention deserves replication in larger samples at other institutions.

We have shown that targeted, specific communication skills can be taught to medical residents in a relatively brief intensive format. Although more comprehensive communication skills training will still likely require a greater time commitment from learners and educators, the intervention we designed may play a useful role for introduction of these skills in housestaff training.


The authors wish to thank Drs. Harvey J. Cohen and Eugene Z. Oddone for their intellectual contributions to the design of this project and their continuous moral support. The authors are indebted to the house officers who participated enthusiastically in the retreats and graciously allowed the observation of their medical interviews.

This project was supported through grants to Dr. Tulsky from the Robert Wood Johnson Generalist Physician Faculty Scholars Program and the Project on Death in America Faculty Scholars Program.


1 Maguire P. Can communication skills be taught? Br J Hosp Med. 1990;43:215–216.
2 Stewart MA. Effective physician-patient communication and health outcomes: a review. CMAJ. 1995;152:1423–1433.
3 Anonymous. Approaching death: improving care at the end of life—a report of the Institute of Medicine. Health Serv Res. 1998;33:1–3.
4 Ramirez AJ, Graham J, Richards MA, et al. Burnout and psychiatric disorder among cancer clinicians. Br J Cancer. 1995;71:1263–1269.
5 Levinson W, Roter DL, Mullooly JP, Dull VT, Frankel RM. Physician-patient communication: the relationship with malpractice claims among primary care physicians and surgeons. JAMA. 1997;277:553–559.
6 Anonymous. Approaching death: improving care at the end of lifed—a report of the Institute of Medicine. Health Serv Res. 1998;33:1–3.
7 Sullivan AM, Lakoma MD, Block SD. The status of medical education in end-of-life care: a national report. J Gen Intern Med. 2003;18:685–695.
8 Fallowfield L, Jenkins V, Farewell V, Saul J, Duffy A, Eves R. Efficacy of a Cancer Res UK communication skills training model for oncologists: a randomised controlled trial. Lancet. 2002;359:650–656.
9 Razavi D, Delvaux N, Marchal S, et al. Does training increase the use of more emotionally laden words by nurses when talking with cancer patients? A randomised study. Br J Cancer. 2002;87:1–7.
10 Razavi D, Delvaux N, Marchal S, Bredart A, Farvacques C, Paesmans M. The effects of a 24-hour psychological training program on attitudes, communication skills and occupational stress in oncology: a randomised study. Eur J Cancer. 1993;29A:1858–1863.
11 Baile WF, Buckman R, Lenzi R, Glober G, Beale EA, Kudelka AP. SPIKES: A six-step protocol for delivering bad news: application to the patient with cancer. Oncologist. 2000;5:302–311.
12 Back AL, Curtis JR. Communicating bad news. West J Med. 2002;176:177–180.
13 Maguire P. Can communication skills be taught? Br J Hosp Med. 1990;43(3):215–216.
14 Quill TE, Arnold RM, Platt F. “I wish things were different”: expressing wishes in response to loss, futility, and unrealistic hopes. Ann Intern Med. 2001;135:551–555.
15 Levinson W, Roter D. The effects of two continuing medical education programs on communication skills of practicing primary care physicians. J Gen Intern Med. 1993;8:318–324.
16 Smith RC, Osborn G, Hoppe RB, et al. Efficacy of a one-month training block in psychosocial medicine for residents: a controlled study. J Gen Intern Med. 1991;6:535–543.
© 2006 Association of American Medical Colleges