While oncologists, patients, and national cancer organizations recognize that communication is important to cancer care,[1,2] exactly how to improve communication has been somewhat less clear. Much of the medical literature addressing communication in cancer care consists of recommendations that reflect expert opinion; recommendations based on empirical research are less common. Yet the existing empirical literature does provide useful, if not quite comprehensive guidance for physicians and other clinicians interested in improving their communication with patients and their families. This article is a selective review of empirical studies about patient-physician communication in oncology, written with the intent of highlighting evidence-based recommendations that can be used by clinicians in practice.
What exactly are communication skills? Communication skills can be defined as the cognitive and emotional abilities used by clinicians to enable patients and their families to understand the nature of the illness, to partner with patients to make sound medical decisions, and to facilitate patient and family adjustment to the illness. While communicating, clinicians engage in intrapersonal processes, such as empathizing or accessing medical knowledge, as well as interpersonal processes, like assessing patient distress or explaining medical information. Some aspects of communication in oncology have been conceptualized as communication tasks, such as giving bad news, that have been used to frame research or education. For example, the trajectory of a cancer illness has been conceptualized as a series of communication tasks, and this approach may be especially useful for teaching.
Since many important communication studies have not used the communication- task heuristic, this review organizes the evidence in the following categories: (1) the patient-physician relationship, (2) how physicians handle medical information, (3) how physicians deal with patient emotions, (4) system-level communication interventions, (5) physician self-management, and (6) educational interventions designed to improve communication.
The distinction between patientcentered and doctor-centered behaviors has emerged from communication research as a principal characteristic about the patient-physician relationship. While the term "patient-centered" has become a quality-improvement buzzword, in communication studies it refers to physicians who use a larger proportion of affective behaviors- such as empathizing or asking openended questions-and actively solicit patient involvement in decision-making (Table 1). In contrast, "doctorcentered" communication refers to physicians who tend to be more biomedically focused and paternalistic, and use more controlling behaviors in the medical encounter. Studies of physician behavior in primary care settings have generally shown patient-centered communication styles to be associated with greater patient satisfaction, better patient recall, and improved adherence to medical recommendations.
Published data indicate that patients and family members prefer a patientcentered approach. An Australian study used videotaped scenarios of an oncologist using a patient-centered or a doctor-centered approach that were viewed by patients with breast cancer, and their family members. The study design called for subjects to separately rate the parts of the consultation involving diagnosis, treatment, and prognosis. The results showed that both patients and their relatives or friends preferred a patient-centered approach in all aspects of the consultation (P < .0001), except in the treatment section of the "good prognosis" video, in which there was no difference.
While this study involved a single site, a single diagnosis, and a high proportion of professional/managerial subjects, it provides compelling evidence of patient and family preferences for a patient-centered approach. Studies have not yet been conducted to examine the association between a patient-centered approach and other oncologic patient outcomes.
An important aspect of the patientcentered approach is in the approach to medical decision-making. Patient-centered does not necessarily equate with high levels of involvement in decisionmaking. Studies of cancer patients have shown that there is a wide spectrum of patient preferences with regard to the degree of involvement they wish to have in decision-making.
For example, in a study of patients with hematologic malignancies undergoing stem cell transplantation, a large proportion of patients (44%) preferred a shared decision-making model in which the physician and patient made decisions together. A significant minority, however, preferred to make the decision themselves, with some input from the physician (14%) or completely on their own (9%). Another significant minority preferred that their physician make the decision, with some input from the patient (17%) or relying strictly on the physician's judgment (18%). The take-home lesson from this study is that a patientcentered approach involves some negotiation about how much the patient wants to be involved in decisions.
While there is evidence that Internet use by cancer patients changes their feelings about their own expertise,[ 6] the only study addressing how this resource influences patient-oncologist communication was a survey of US oncologists. Respondents reported that Internet use made patients more hopeful, confused, anxious, and knowledgeable-simultaneously.
Finally, empirical evidence suggests that a physician's caring attitude is a therapeutically important aspect of the relationship. In a study of bad-news conversations in which women with breast cancer were surveyed 6 months after diagnosis, the physician's caring attitude was the strongest predictor of adjustment to illness. Another study asking patients to compare two videotapes showed that 40 seconds of compassionate behavior was associated with lower levels of patient anxiety. Evidence-based recommendations for the patient-doctor relationship are listed in Table 2.
Handling Medical Information
A large number of survey studies of cancer patients indicate that overall satisfaction with communication is quite high. This probably reflects both the ceiling effects of satisfaction instruments, and the relative lack of patient experience on which comparative judgments about physician communication skills can be made, since most patients have significant contact with only a single oncologist. While satisfaction may not have changed much historically, the practice of information disclosure has changed dramatically in the United States over the past 40 years, with patients now provided much more in terms of specific diagnostic and prognostic information.[ 10] In a British survey of 2,331 cancer patients, 98% indicated that they wished to have "all information" about their cancer. Of note, however, 5.4% said that they wanted "only good news."
Cultural influences on patient desire for explicit information disclosure are substantial. In a survey of 800 seniors in Los Angeles, 87% of Caucasian Americans believed that a patient should be told about a diagnosis of metastatic cancer, as compared to only 35% of Korean-Americans. Thus, while it is fair to assume that most patients want to receive explicit and complete medical information about their illness, a notable minority do not. Although guidelines have been suggested for dealing with cultural differences, there are no empirical studies that describe how physicians handle requests for limited information disclosure in oncologic settings.
There are notable gaps between the information patients say they would like to have received and what they did receive. In a survey of patients with breast cancer or melanoma, 57% wanted to talk about prognosis, yet only 27% actually did; likewise, 63% wanted to talk about the effects of the cancer on other aspects of their life, yet only 35% did. In general, oncologists tend to focus on anticancer interventions to the exclusion of other issues. In studies of patients with advanced cancer in which a research nurse used a checklist to elicit patient concerns after a visit with an oncologist, patients were found to have numerous concerns they had not disclosed to their oncologists, and a larger number of concerns was correlated with poorer adjustment to illness. These data suggest that oncologists should spend some time during each visit eliciting patient concerns.
Giving Bad News
Giving bad news is the most highly studied communication task in oncology. Patients describe a variety of reactions-including shock, fright, sadness, and even acceptance-to receiving a diagnosis of cancer. A large survey found broad agreement on a number of principles for giving bad news between patients and clinicians,[ 16] and these have been incorporated into specific protocols.[17,18] A smaller focus group study found that patients and family members with advanced cancer wanted physicians to be honest and straightforward but also sensitive in giving bad news. Although a number of protocols for giving bad news have been designed and successfully taught in educational interventions, no studies have evaluated the importance of individual steps of a protocol, compared the use of different bad news protocols, or linked them to patient outcomes such as distress or anxiety after the bad news visit.
There are only limited data examining the best way for oncologists to present and help patients evaluate treatment options. For example, 43% of patients with breast cancer preferred a positively framed message (ie, discussing "chance of cure"), while 33% preferred a negatively framed message (ie, discussing "chance of relapse"). Clearly, many patients lack an understanding of how to use statistical information such as median survivals, and a substantial number of patients have difficulty comprehending and retaining numerical data. In a study in which patients responded to clinical vignettes, the absolute survival benefit was shown to be the preferred method of discussing the benefits of undergoing adjuvant chemotherapy for breast cancer. Taken together, these data suggest that clinicians should consider "double-framing" statistical data-ie, explicitly naming both the chance of cure and the chance of relapse, and discussing absolute survival benefits rather than relative risk reduction.
Studies show that after discussions about clinical trials, patients and family members commonly misunderstand fundamental concepts. Although one of the top reasons patients cite for enrolling in a clinical trial is "trust in the doctor," it is clear that trust does not necessarily translate into understanding.
A study of patients entering phase I trials showed that 85% had decided to participate because of possible therapeutic benefit, and only 33% were able to state the purpose of the trial.[ 24] In a multicenter study of childhood leukemia trials in the United States, postconsent interviews showed that 50% of parents did not understand randomization. A survey of 600 patients in the UK comparing preferences for different explanations of randomization used by major cancer organization websites and information sheets showed that only 54% to 61% of patients rated the explanations as "very clear," and only weak consensus could be found as to which explanations were most preferred and disliked. These data suggest that oncologists cannot rely on a single type of explanation and must check a patient's understanding, with the possible necessity of reexplaining concepts using a different frame of reference.
Studies examining how physicians disclose prognosis indicate that accuracy is variable and that a substantial proportion of oncologists are reluctant to give any kind of prognosis. In a prospective cohort study of 313 patients undergoing autologous and allogeneic stem cell transplants, physicians tended to overestimate survival, especially for poor-prognosis patients, while patient expectations remained high regardless of disease stage. At the other end of the illness trajectory, physicians are similarly inaccurate. In a study of patients entering hospice, only 20% of physician predictions were accurate and 63% were overoptimistic.
In another study, oncologists reported that they would not provide any survival estimate 22% of the time, would provide their own survival estimate 37% of the time, and would provide a different survival estimate (usually optimistic) 40% of the time. This may be because it is difficult to give patients a prediction that their life expectancy is limited, or because oncologists want to support patient hope, but the reasons have not been extensively studied.
An in-depth qualitative study of oncologists at Harvard teaching hospitals found that these oncologists consciously participated in "a discourse about hope" that had to do with life prolongation, the promise of biomedical technology, and the use of outliers as examples. In another interview study, oncologists used the metaphor of "hitting you over the head" to describe the impact of prognostic disclosure on patients. In the hospice study mentioned above, as the duration of the patient-doctor relationship increased, prognostic accuracy decreased. This may speak to the difficulty of confronting patients with the prospect of dying.
Palliative Care Issues
Empirical studies examining communication about palliative care issues, such as advance care directives, do-not-resuscitate orders, and discussions about withdrawal of life-sustaining therapy highlight a number of deficiencies. Such investigations have often involved cancer patients but usually involve nononcologists-residents, primary care physicians, or critical care intensivists. These studies indicate that physicians tend to focus narrowly on technical decisions, fail to elicit patient values, and do not clearly discuss the likelihood that the life-sustaining intervention will meaningfully prolong life. Most patients with incurable cancer prefer a realistic and individualized approach.
A secondary analysis of a large unsuccessful intervention study that provided detailed prognostic information to hospitalized patients with a 50% 6-month prognosis showed that patients whose estimates of prognosis were much more optimistic than their physicians were more likely to die after an attempted resuscitation or mechanical ventilation, suggesting that patient misperception led to decisions to pursue life-sustaining therapy that ultimately proved unsuccessful.
One interview and focus group study examined the question of when and how to initiate discussion about prognosis and end-of-life issues, and concluded that physicians should clarify the patient's understanding and how much detail he or she wants to know, as well as negotiate who should be present and who should deliver the information.
One empirical study focusing on the transition to hospice care, involving a survey of bereaved family members of cancer patients who had died in an inpatient hospice unit, indicated that a substantial proportion felt that communication needed to be improved and that family distress correlated with the family being told that "nothing more could be done." Studies addressing the components of "a good death" that included cancer patients identified six process-oriented themes: pain and symptom management, clear decisionmaking, preparation for death, completion, helping others, and affirmation of the whole person.
Taken together, these studies suggest that improving communication about palliative care for cancer patients should involve a discussion of patient values and a clear statement about nonabandonment with specific care plans (rather than a statement that "nothing more can be done"), and that physician knowledge of the elements of a good death may help guide a productive conversation. No studies have yet examined the effect of instituting these communication practices on patient outcomes. Evidencebased recommendations for handling information are listed in Table 3.
Dealing With Patient Emotions
The skills required for a clinician to deal effectively with patient emotions, especially distress, involve detecting and identifying patient emotion, responding to the emotion with empathy, and in some cases, further medical assessment or intervention for distress or mood disorders. Studies of psychological morbidity in cancer patients indicate that between 35% and 50% have psychological problems,[ 37] with the most common diagnosis being adjustment disorder with depressed mood. Many cases of distress, however, are undetected by clinicians and therefore go untreated.
One large British study involving 143 oncologists and 2,297 patients in outpatient encounters indicated that the physicians' sensitivity in detecting psychiatric morbidity was only 28%, and the misclassification rate was 34%. In another study of oncologists' perceptions about clinical depression, only 6% of patients who were clinically depressed as measured by a well-validated instrument were detected as depressed by their oncologist.
Patient and Physician Attitudes
Why is the rate of clinician detection of significant distress so poor? It appears from the literature that the answer lies in patient and physician attitudes and their effect on the ability to elicit patient clues. A variety of patient attitudes have been identified as barriers to communication about emotional distress. Many patients perceive that their physicians are too busy or not interested in hearing about emotional concerns such as distress. For example a British study of primary care practices indicated that 45% of patients experiencing high levels of distress did not disclose their symptoms because they felt embarrassed or hesitant to bother the physician, and 19% felt deterred by the doctor's behavior during the visit.
In addition, patients may prioritize informational issues. In an Australian study examining audiotapes of cancer consultations, patients gave more verbal cues relating to informational needs than emotional needs. While patients with more education levels are more likely to be willing to present emotional concerns, it has not been shown that age or gender reliably predict patient willingness to disclose emotional concerns.
Physician attitudes are also responsible for poor communication about emotional issues. Many clinicians assume that patients will request help for a significant problem, and therefore they do not probe for information about emotional issues, deferring to the patient to initiate discussions about psychological issues.[40,43] Physicians may also feel uncomfortable about handling discussions involving strong emotions, and one study indicates that physicians experience significant stress when disclosing bad news to patients. Oncologists who lack communication skills training report the most anxiety and the least confidence when dealing with emotional issues.
Despite clinician attitudes suggesting that they do not have the time to deal with emotional issues, an audiotape study indicated that clinician responses to emotional cues are associated with shorter visits. Finally, studies done in primary care settings indicate that physicians who blame patients for causing or exaggerating symptoms are less likely to detect emotional distress. That said, clinicians who have a positive attitude toward psychosocial concerns but lack communication training are not better at eliciting emotional concerns, at least in primary care settings.
Two Levels of Communication
Thus, the literature shows that detecting and responding to emotion is a separate communication skill from dealing with informational needs. It may be helpful to consider communication skills on two levels: informational and emotional. Some data indicate that strong emotions interfere with informational processing. In one study of patients undergoing radiotherapy for cancer, investigators found a significant correlation between anxiety and depression scores and dissatisfaction with the information received.[ 48]
Simply providing more information when a patient is distressed appears not to be useful, and in fact may worsen distress. In a randomized study of patients who were provided with audiotapes of their visits with an oncologist, a secondary analysis indicated that poor-prognosis patients had more distress at 6-month followup.[ 49] Thus, simply replaying the visit in the absence of active supportive care may be actively unhelpful for these patients.
Finally, a study in which initial and follow-up visits to a hematology outpatient clinic in Norway were audiotaped showed that patients were more satisfied when they were able to express emotional concerns in followup visits and physicians responded to their emotional concerns (but not during initial visits), suggesting that once informational needs are met, relationship- building through discussion of emotional issues becomes more important.[ 50] Evidence-based recommendations for dealing with patient emotions appear in Table 4.
System-Level Communication Interventions
The Institute of Medicine National Cancer Advisory Board concluded that communication was suboptimal for many patients and noted a number of system characteristics that hinder communication. Some of these characteristics are not mentioned elsewhere in this paper, including too little time per patient, consulting room layouts that make small group conversations cramped or difficult, and reimbursement of oncologists for time needed for communication. None of these factors have been subjected to empirical study.
A number of system-level communication interventions have been described in the literature, including written treatment plans completed during the visit, asking patients to write lists, audiotapes of the visit, question prompt sheets, and dictated summary letters sent to the patient after the visit. Audiotapes and question prompt sheets have been studied in randomized controlled trials. One study showed that at 6 months of follow-up, the patients who received audiotapes recalled significantly more information about their illness, that their psychological distress at 1 and 6 months was no different than that seen in controls, and that for poor-prognosis patients receiving audiotapes, psychological distress at 6 months was worse.
A thoughtful intervention package studied in Australia, consisting of a question prompt sheet, a booklet on clinical decision-making and patient rights, and a booklet introducing the clinic, was compared in a randomized trial to use of the introduction-to-theclinic booklet only. This study showed that the patients receiving the package were significantly more anxious, asked more questions, and challenged physician information significantly more often. Yet the intervention patients were less likely to achieve their preferred decision-making style than controls (22% vs 35%). This study underscores the importance of physician engagement in a system-level communication intervention.
Providing physicians with qualityof- life information independently col- lected from the patient was shown in a randomized controlled trial to increase discussion and physician awareness of quality-of-life issues. This system intervention, although the end point was discussion of quality-of-life issues rather than improvement in patient quality of life, is the system-level intervention best supported by empirical evidence.