Discussing prognosis and treatment goals with patients with advanced cancer: A qualitative analysis of oncologists’ language

Abstract Background The National Academy of Medicine recommends that cancer patients be knowledgeable of their prognosis to enable them to make informed treatment decisions, but research suggests few patients receive this information. Objective This qualitative study describes oncologists’ language during discussions of prognosis and treatment goals in clinical interactions with African American patients diagnosed with cancer. Design We analysed transcripts from video recordings of clinical interactions between patients with Stage III or IV cancer (n=26) and their oncologists (n=9). In‐depth discourse analysis was conducted to describe and interpret oncologists’ communication behaviours and common linguistic features in the interactions. Setting and participants Data were from a larger study of patient‐provider communication between African Americans and oncologists at two cancer hospitals in Detroit. Results Prognosis was discussed in 73.1% (n=19) of the interactions; treatment goals were discussed in 92.3% (n=24). However, analysis revealed that oncologists’ description of prognosis was vague (e.g. “prognosis is a bit worse in your case”) and rarely included a survival estimate. Oncologists often used ambiguous terminology, including euphemisms and jargon, and emphasized uncertainty (e.g. “lesions are suspicious for the disease”). Conversation about prognosis was frequently brief, moving quickly to the urgency and details of treatment. Discussion This study demonstrates how oncologists’ language may obscure discussion of prognosis and treatment goals. The identified behaviours may lead to missed opportunities in eliciting and discussing patients’ knowledge about and preferences for their care. Patient‐, provider‐ and system‐oriented interventions are needed to improve clinical communication, especially among minority patients with advanced cancer.


| INTRODUCTION
A recent report by the National Academy of Medicine concluded that cancer care delivery in the United States is in crisis. 1 The care that patients receive is often not as patient-centred, accessible or coordinated as it could be. Many of these problems apply especially to patients with advanced cancer. Numerous recent studies have shown that aggressive treatments continue to be widely administered to patients who are not adequately informed of the status of their cancer or the goals of their treatment. Moreover, despite the well-documented benefits of palliative care-from improved symptom management and psychosocial outcomes to overall quality of life, and even longer survival time-there is a pervasive underutilization of palliative care for advanced cancer patients. [2][3][4][5] This underutilization is greater among minority individuals, relative to White patients. [6][7][8][9] Among the myriad reasons for suboptimal care provided to patients with advanced cancer, poor-quality patient-provider communication is recognized as a major factor. [10][11][12][13] Indeed, despite the National Academy of Medicine's recommendation that cancer patients be informed of their prognosis to enable them to make informed treatment decisions, few patients actually receive this information. 1,13 When such conversations do occur in clinical encounters, they have been observed to be problematic in that they do not achieve the goal of patient understanding of prognosis. For example, in a UK study with patients participating in Phase I Clinical Trials, Jenkins et al. 13 observed that prognosis was only discussed in 21% of conversations (n=52) and physicians rarely checked patient understanding of prognosis. Compounding the difficulties in communication about prognosis, patient-provider communication tends to be worse (e.g. less information exchanged) for ethnic/racial minority patients, including African Americans in the United States. 6,9,[14][15][16][17] Similar patterns hold for communication at the end of life. 6,18 Poor communication quality may be a factor in the well-documented health disparities in cancer treatment and outcomes. 16,19,20 Hence, in-depth investigation into the way prognosis and goals of care are communicated during clinical interactions with African American patients with advanced cancer is especially warranted.
Recognizing the central role of patient-provider communication in cancer care, social scientists over the last decade have endeavoured to better understand oncology prognosis discussions-including their barriers and facilitators-using a variety of approaches. For example, Leydon (2008) qualitatively analysed oncology consultations about cancer treatment options, focusing on the role of conveying optimism in light of uncertainty or "bad news". 21 Many similar communication studies have been done on prognosis conversations. For example, in an observational study of clinical interactions, Robinson et al. analysed transcripts of 141 interactions between oncologists and patients with advanced cancer. 22 They found that while treatment was discussed in the majority of interactions (94%), prognosis (defined as "a statement about expectations of the disease that refer to the likely course of the patient's cancer or what the outcome might be") was discussed in only 50% of the interactions. They also examined features of the oncologists' communication behaviours that correlated with patient-provider concordance in understanding of the chance for a cure: when oncologists made at least one pessimistic statement, patients were more likely to agree with their oncologist's estimated chance of cure. On the other hand, statements of optimism and uncertainty were not associated with an increased likelihood of concordance. Finally, in the context of medical interviews in which a patient's death and dying are discussed, a Conversation Analysis project described the contribution of the patient's responses to the oncologists' delivery of the news about patients' poor prognosis. 23 In addition to observational studies, researchers have analysed patient interviews and other self-reported data in order to better understand how patients are informed about prognosis during clinical interactions. For example, Step and Ray conducted in-depth interviews with 30 female cancer patients about their experience with prognosis conversations during initial diagnosis and cancer recurrence. 24 Patients reported that initial diagnosis conversations included physicians communicating optimism about prognosis, but during recurrence, the conversations shifted to a focus on the logistics of disease management.
During recurrence conversations, much more uncertainty was communicated, leading to tension-filled "prognosis dance" where patients and physicians both seek and avoid information 24 (p. 54-55). Finally, palliative care practitioners and educators have attempted to describe key problems in prognosis discussions and offer practical guidelines to improve discussions in clinical practice. 10,25,26 This paper builds upon prior observational research by providing an in-depth examination of oncologists' language in real-life medical encounters with African Americans with advanced cancer.
Discourse analysis, a qualitative approach to analysing interactions, is useful in this context because it is grounded in theories derived from linguistics, communication science, psychology and sociology and may help researchers identify common behaviours employed by oncologists during sensitive encounters and to interpret and explain these behaviours. [27][28][29][30] This analysis serves to further explain factors that may contribute to patients' poor understanding of the status of their disease. The findings from this research can inform the assessment of oncologist communication, as well as the development of interventions. The present analysis aims to describe oncologists' communication behaviours and language during initial discussions about prognosis and treatment goals with a small sample of African Americans with advanced breast, lung, or colorectal cancer.

| METHODS
Study Data were taken from a larger National Cancer Institute-funded study conducted from 2012 to 2014, designed to improve patientoncologist communication during racially discordant interactions in the outpatient clinics of two cancer hospitals in Detroit, Michigan, USA. 31,32 Oncologists were eligible to participate if they treated patients with breast, colon or lung cancer. Patients (n=137) participated in the larger study if they self-identified as Black, African American or Afro-Caribbean, were between the ages of 30 and 85, were able to read and write English well enough to provide consent and answer the questionnaires, had a confirmed diagnosis of breast, colorectal or lung cancer and were scheduled for an appointment to see a medical oncologist to discuss medical treatment (e.g. chemotherapy). Thirtyfive of the 40 eligible oncologists (87.5%) agreed to participate, and of these, 18 (51.4%) had patients who participated and were included in the study sample.
The Institutional Review Boards of both hospitals and Wayne State University approved all procedures. Data from the present study consisted of transcripts of video recordings from encounters with all patients with Stage III (n=20) or IV (n=6) cancer, their family caregivers (when applicable), and their oncologists (n=9), as well as participants' basic demographics and patient's medical information. These 26 encounters comprised the study sample. Mean length of encounters was 36.59 minutes (SD=15.74). Table 1 describes the participant characteristics.
The following steps were followed to obtain study findings. First, our multidisciplinary team of investigators, including behavioural scientists and oncologists, reviewed all 26 transcripts to reach consensus on definitions of "prognosis" and "treatment goals" discussions.
Prognosis Discussion was operationally defined as any mention of the probable course or outcome of a disease, especially the chances of recovery. Goals of Treatment Discussion was operationally defined as any mention of why a treatment is being considered (e.g. to cure, treat symptoms or shrink the tumour). Two authors (LH, SE) then independently identified and extracted transcript excerpts in which discussions of treatment goals and prognosis occurred; the selected excerpts were then verified by two other authors (WSC and CT). Disagreement among team members was resolved through discussion.
Focusing on the extracted excerpts, we performed a qualitative discourse analysis, with the goal of identifying oncologists' communication behaviours, as well as specific linguistic features related to prognosis and goals of treatment. 27 The first author, a trained sociolinguist, identified prominent features that emerged and offered interpretations informed by theories of social interactions. 27,29 She and the study team then discussed preliminary findings and refined interpretations through an iterative process. In presenting the following qual-

| Frequencies of prognosis and goals of treatment discussions
Prognosis discussions occurred in 19  news," referring to the treatment he is proposing. Such brief mention of prognosis followed by other information is typical of language used by oncologists in the study sample.
In contrast to prognosis, treatment goal discussions occurred in most (n=24, 92.3%) of the interactions:

Example 3 (Pt C; Stage IV Breast) So what the chemotherapy does is that
it tries to attack those cells and prevent the cancer from coming back.
In this example, the oncologist explains how chemotherapy works and what its goals are, as part of the discussion of a treatment plan.
Such discussions occurred frequently in the sample. The next section will explore key strategies and features commonly observed, with the goal of shedding light on how these sensitive discussions are managed or minimized.

| Oncologists' discussion of prognosis and treatment goals: Macrolevel communication strategies and micro-level linguistic features
Our analysis was informed by a discourse analytic framework, which allowed us to distinguish between macro-level communication strat- (1) euphemism/ambiguity, (2) modal expressions and (3) jargon. Each of these observed strategies is summarized below, beginning with a description and definition, followed by transcript illustrations and interpretations based on relevant theories.

| Macro-level communication strategies
(1) Minimizing prognosis discussion by moving quickly to treatment goals/plans or merging prognosis with goals of treatment talk. This observed sequence of topics differs from the typical "routine" structure of medical encounters, as described by Ten Have's conversation analysis of medical consultations: opening->complaint->examination->diagnosis->treatment/advice->closing. 30 Ten Have posits that a diversion from this typical sequence is marked as it runs counter to the expectations of the participants. Upon mentioning that the cancer has grown, the oncologist leaves no time to discuss the cancer diagnosis and its implications for a poor prognosis. Instead, he launches into the next topic-the urgency of treatment. This type of switch from referencing a poor prognosis directly to treatment planning is very common in these data, whereby oncologists either move quickly from "diagnosis" talk into "treatment/advice" talk or merge the two.

(2) Emphasizing uncertainty and lack of information
In discussing prognosis and goals of treatment, oncologists were observed to emphasize the fact that they did not have sufficient information about the patient's disease. Example 5 illustrates a typical discussion focusing on future tests and information gathering:

(3) Emphasizing optimism and hope
Oncologists were observed to emphasize hope when a potentially poor prognosis was indirectly referenced: This oncologist referenced a poor prognosis, namely, the limited chance of a cure. However, right after admitting "being candid" about the prognosis, he established his belief in "doing this" -what "this" entailed was unclear, but a sense of optimism in the treatment was conveyed. Expressions of optimism in patient-provider encounters have been studied. For example, Robinson et al. examined the association between physicians' expressed optimism and patients' and providers' alignment in prognostic understanding. 22 The researchers found that physicians' expression of optimism is linked to misaligned understanding of prognosis. It is possible that a focus or orientation towards possible effective treatments and a sense of optimism, though well intentioned, contribute to confusion or misunderstanding about the diagnosis and prognosis.

| Micro-level Linguistic features
How are the above-mentioned communication strategies achieved linguistically? A discourse analysis identified three micro-level linguistic features; each of which will be described and defined below, along with illustrative examples and a theory-based interpretation. (

1) Use of euphemism and linguistic hedging
Oncologists were observed to broach the topic of prognosis indirectly. One common way of signalling indirectness is through the use of euphemistic expressions. Euphemisms are a kind of linguistic hedging which functions to signal the speaker's sense of uncertainty towards a proposition and downplay their epistemic stance towards a statement. 28 For example, words such as "lesion," "spot" and "stuff" were used frequently to reference cancer, and a generic and technical term, "activities," was used to reference cancer growth.

Example 7 (Pt I; Stage III Lung) So I want to make sure it's abnormal tissue. It's a mass or lesion or tumor or…
The oncologist brings up a number of alternative terms-from "abnormal tissues" to "mass," "lesion" and "tumour" -none of which offers a definitive diagnosis or prognosis, adding to ambiguity and potential for confusion. In addition to these generic referring terms for cancer, there are other euphemistic expressions in the language of the oncologists. from the discussion, one might infer that it is inoperable due to the size. The generic noun "everything" indirectly refers to the cancer.
Note also that the verb "treat" here does not entail a curative measure, but merely helps to reduce the size of the tumour to make surgery possible as a subsequent treatment option. Finally, the phrase "improve your chances" again indirectly suggests a grim prognosis, and this indirectness is further shown with linguistic hedges such as "could" and "try to." The discourse marker "you know," uttered twice by the oncologist solicits the patient's affirmation of the receipt of previously mentioned information and possibly signals the speaker's sense of uncertainty towards the proposition. 29 Linguistic hedging serves to convey ambiguity; for example: Example 9 (Pt L; Stage III Lung) The hope is that the total package gets rid of this…but whether we will or not only time will tell.
In discussing the goal of the treatment and its potential effectiveness, the oncologist's language sets up "hope" but then leaves the prospect of getting rid of cancer uncertain with the phrase "only time will tell." This language could potentially leave the patient confused about the goal of treatment and likelihood of it being effective.
Taken together, the frequent use of euphemisms and hedging to discuss prognosis suggests that oncologists may find that directly addressing this sensitive topic to be face-threatening, and hence mitigate this threat through language. This explanation would align with the framework of face 34 and politeness theory. 35 These theories posit that in social interactions, including and particularly face-to-face exchanges, in addition to conveying information, speakers have to constantly mitigate potential "face threats" (i.e. threats to a desired self-image) carried by certain face-threatening acts towards another individual. Talking to a new patient directly about a life-threatening diagnosis and their mortality is inherently face-threatening, and naturally oncologists find strategies to mitigate the uncomfortable discourse, often through indirect speech.

(2) Use of Modal expressions
Indirectness in the oncologists' language can also be observed by their use of modal expressions such as conditionals, especially in the discussion of prognosis. In response to the family member's question about prognosis, the oncologist suggests that fifty percentage of patients with Stage IV lung cancer will live more than two years, followed by the use of a series of conditional expressions ("if…") to discuss different possible prognostic outcomes depending on whether the cancer is only in the bones or also in the lungs. The discussions are inherently modal: modality in linguistics refers to "expressions that relate to potentially unreal situations". 36 It is impossible for the patient and doctor here to know exactly whether the cancer has metastasized to the bones. Bringing out alternative scenarios through modal expressions contributes further to uncertainty and confusion about the prognosis, which is typically poor for a Stage IV diagnosis.
Later on in the same encounter, the doctor continues the discussion of the prognosis in Example 11 through conditional expressions ("If…") and hedges ("we think, we believe"). Here, the oncologist touches upon the potential of dying from cancer, but minimizes the possibility by providing many hypothetical scenarios. This is typical, wherein oncologists used modal expressions and listed alternative prognostic scenarios (some of which were unreal) as a way to minimize the potential for a poor prognosis.

(3) Use of Complex language and medical jargon
Use of complex language and medical jargon were observed fre- In addition to descriptions such as "these lesions are kind of suspicious for the disease," the complex language use potentially causes more ambiguity and confusion during the discussion of high-stake topics including prognosis and goals of treatment.  Methodologically speaking, our qualitative discourse analysis complements other observational and survey methods frequently used to study patient-provider communication. A focus on language allows us to dive deeper into the sequential unfolding of interactions and provides partial explanations for why these exchanges often lack key aspects of the prognosis and goals of treatment. In essence, we observe a common trajectory of interaction focusing on future events that may be unlikely to provide a cure.

| DISCUSSION
The limitations of this study include the fact that even though the interactional context is taken into account in our qualitative study, the central focus is on the oncologist and their communication behaviours.
In addition, we only analysed transcripts; therefore, we were unable to account for non-verbal behaviours in these encounters. The dynamic, interpersonal process of clinical communication about prognosis warrants further investigation. 49 Future research should conduct dyadic and triadic analyses on these interpersonal communication processes, as well as focusing on patient behaviours. Another limitation is the fact that this is a small sample taken from one patient population from one geographic location, and thus, findings may not be generalizable.
However, this is one of the first studies to conduct this type of indepth, descriptive analysis of clinical interactions with African American patients. This population bears a disproportionate burden of cancer disparities including incidence, mortality, communication and care. 20,50

| CONCLUSION
This study examined oncology communication in action through an in-depth qualitative linguistic analysis of real-life, real-time clinical encounters with patients diagnosed with advanced cancer. Through discourse analysis, we identified a number of oncologists' communication behaviours and described how they fit into the overall consultation genre, 30 in order to better understand how sensitive topics such as prognosis and goals of treatment are (or are not) discussed.
This type of in-depth analysis complements counting/coding studies, allowing for a description and interpretation of not only what is dis- cussed, but what is not discussed. In addition, the study points to next steps in research, including in-depth interviews with patients, caregivers and providers to complement observations and triangulation across multiple sources of data (e.g. care utilization data or secondary surveys). With rapid advances in cancer treatment and promising new drugs being developed, it is likely that prognostic information will only become more confusing and uncertain. This study has the potential to inform communication training programmes, helping providers to be more mindful of their language use in order to effectively convey high-stake information such as prognosis and goals of treatment to their patients.