Getting on the same page: Communication, patient involvement and shared understanding of “decisions” in oncology

Abstract Background Patients and clinicians do not often agree on whether a decision has been made about cancer care. This could be explained by factors related to communication quality and/or the type of decision being made. Methods We used a self‐developed coding scheme to code a random sample of 128 encounters in which patients and clinicians either agreed (n=64) or disagreed (n=64) that a cancer care decision was made and tested for associations between concordance and key communication behaviours. We also identified and characterized cancer care decisions by topic and level of patient involvement and looked for trends. Results We identified 378 cancer care decisions across 128 encounters. Explicit decisions were most commonly made about topics wherein decision control could be easily delegated to a clear and present expert (eg either the patient or the clinician). Related to this, level of patient involvement varied significantly by decision topic. Explicit decisions were rarely made in an observable way about social, non‐clinical or self‐management related topics, although patients and clinicians both reported having made a cancer care decision in encounters where no decisions were observed. We found no association between communication behaviours and concordance in our sample. Conclusions What counts as a “decision” in cancer care may be constructed within disparate social roles that leave many agendas unaddressed and decisions unmade. Changing the content of conversations to encourage explicit decisions about self‐management and life context‐related topics may have greater value in enabling shared understanding than promoting communication behaviours among already high‐performing communicators.


| Background
Optimizing patient-clinician communication is a priority intended to improve the quality of care in oncology and positively impact patient quality of life, patient satisfaction and medical outcomes. [1][2][3][4][5] Shared decision-making approaches that increase patient involvement in the decision-making process are particularly promoted because these approaches seek to ensure that the outcomes of care-related decisions reflect the values and preferences of patients. In previous research we showed that, when surveyed immediately after an outpatient oncology encounter, patients and clinicians often disagreed in their ratings of whether any "specific cancer care decision was made" within that encounter. 6 This study raised questions about the quality of decision making in oncology and highlighted the possibility that patients and clinicians may define "cancer care decisions" differently. 7 It also suggested a potential link between decision making and a broader indicator of poor quality communication long recognized in oncology, namely discordance between patients' and clinicians' understanding (this has been seen in regards to diagnosis, prognosis and treatment preferences, among other things). [8][9][10][11][12][13] To date, the underlying and causal communication failures that contribute to discordance in oncology are poorly understood, and our previous work shed no additional light. 6 Identifying and describing these factors could be of value in avoiding discordance and improving communication, decision making and care quality in oncology.
Some evidence suggests that explicit communication techniques may be needed to ensure patients and clinicians remain "on the same page" and that patient-centred goals for life and health are pursued. [14][15][16][17] Studies have shown, for example, that physicians often do not take the time to fully establish their patient's agenda or goals for an encounter. 18,19 In multiple medical contexts, this has been shown to lead to poor understanding of the patient's problem and desires for care by the physician. [20][21][22] It is possible that, if patients and clinicians have different expectations of or agendas for an oncology encounter, these might influence their perceptions of what was or was not accomplished in it, and thus lead to discordance. Patients and clinicians could also have different perceptions of whether a "cancer care decision was made" based on their level of involvement in the decision-making process itself. In that sense, discordance might not only be expected when patients and clinicians have different ideas about what cancer care decisions are, 7 but also when they make independent and undisclosed decisions about their own intentions for "cancer care." These complex issues have not been explored.

| Objective
The objective of this study was to identify possible factors that could predict patient-clinician concordance in perception of whether any cancer care decisions are made during oncology encounters. More specifically, we sought to assess the effects of clinicians' efforts to, (i) elicit the patient's encounter agenda, (ii) summarize the activities of the encounter and, (iii) assess the patient's understanding of the plan. We also categorized all cancer care decisions identified by both topic and level of patient involvement and looked for trends and associations.

| Sample and participants
Patients and clinicians in the encounters were enrolled in a parent study. 23 Eligible patients were at any stage of management for any solid tumour malignancy. They were being seen in their regular outpatient encounter at a large, tertiary cancer centre in the upper Midwest of the United States. Clinicians were staff oncologists, senior fellows or nurse practitioners.

| Procedure
This study is a follow-up analysis of our 2015 study on patientclinician discordance. 6 In that study, we used survey data to identify situations in which patients and clinicians agreed (eg were concordant) or disagreed (eg were discordant) in their assessments of whether a cancer care decision was made in an outpatient oncology encounter. In the concordant sample (n=213), patients and clinicians either both responded "yes" or both responded "no" when asked if a decision was made. In the discordant sample (n=102), one responded "yes" when the other responded "no" to the same question (note that this was a "best-case scenario" evaluation of concordance as we could not guarantee that patients and clinicians were identifying the same decisions). Surveys were collected immediately after the encounter. The encounters were audio recorded as part of the parent study, but we did not access audio recordings at the time of the 2015 study.
For this analysis, our statistician used a random sequence generator to select a random sample of 64 encounters each from both the concordant and discordant samples of the 2015 study (128 encounters in all). We estimated, using the effect size of a similar study in primary care, 22 that this sample would give us 80% power to detect a meaningful difference in the primary outcome of whether eliciting the patient's agenda is associated with concordance at 95% confidence with a two-sided alpha.
Recordings of the encounters were transcribed verbatim and uploaded to a web-based data system for coding (REDCap 24 ). Our research team iteratively reviewed pilot transcripts that were not part of the study sample and consulted experts to inductively develop the coding criteria for all outcomes. After calibrating judgements on the pilot sample, a team of two analysts (AL, physician researcher and cancer patient; JH, qualitative analyst experienced in patientclinician discourse) coded all study transcripts, while blinded to their concordance status. They coded the first 20% of the sample (26 encounters) in duplicate. After establishing interrater agreement at >75% for all codes, AL coded the remaining 80% of the encounters alone.

| Measures
For the primary outcome, coders took note of whether the clinician made an effort to "elicit the patient's agenda." Possible ratings were "clearly" (eg "What would you like to get out of our time today?") "partially" (eg "What concerns or questions do you have?") and "not at all" (eg "How are things going?"). Ratings were of clinician behaviour primarily but were outcome focused. In that sense, it was possible for encounters to be coded as "clearly" or "partially" if patients were judged to "volunteer" their agendas to that degree.
To explore other communication behaviours that might account for concordance, coders similarly took note of (i) the extent to which the clinician confirmed the course of action with the patient by restating it at the end: possible ratings were "clearly" (eg "Let me summarize what we decided on today…") "partially" (eg "So we'll switch to drug X and see you back in…") and "not at all" (eg "Okay, any questions for me?"); and (ii) the extent to which the clinician assessed the patient's understanding of what occurred: possible ratings were "clearly" (eg "Can you tell me what we decided to do today?") "partially" (eg "Is everything clear? Do you have questions?") and "not at all" (eg "Let's proceed with this plan then.").
Simultaneously, coders flagged every instance in which a cancer care decision was made according to a set of pre-developed criteria, and categorized each decision by topic and degree of patient involvement in making the decision (see Table 1 for details and categories).
We developed all coding criteria after a thorough literature search and consultation with experts. We confirmed that no existing measures were feasible or appropriate for this study, often because they targeted different constructs or used conceptualizations of medical decisions that were too narrow.

| Statistical analysis
We entered all coded data from the encounters into JMP statistical software and linked it to the concordance status as recorded in the 2015 study. Only rarely did we observe "clear" and explicit examples of the communication behaviours we were assessing. As such, we elected to dichotomize all of the communication behavioural outcomes by combining ratings of "clearly" and "partially." We then used chi-square tests of association to evaluate the effects of clinician behaviours, setting concordance as the dependent variable. We summarized distributions with descriptive statistics and explored the data with bivariate tests of association.

| Sample characteristics
Concordant and discordant samples were well balanced with regard to participant characteristics (Table 2). Overall, patients were mostly Was NOT information-sharing, advice-giving, or history-taking 8. A decision to pursue clear action that will lead to a decision will count (ie determine insurance and then call back with decision, etc.)

| Discussion
We sought to explore the effects of patient-centred communication behaviour and decision-making factors on the development of shared understanding in oncology. We were unable to detect an association between any of the communication behaviours we evaluated and patient-clinician concordance in the assessment of  Aside from this, this study represents, to our knowledge, one of the largest and most rigorous analyses of patient-centred communication and decision making in oncology.

| Conclusion
In summary, we showed that explicit decisions were often only made in relation to topics where there appeared to be a clear and present expert (either the patient or the clinician). We also showed that pa-

| Practice implications
The importance of communication in oncology has long been recognized. 28 Although high quality patient-clinician communication is accepted as a central component of patient-centred care, the exact ways in which it is able to contribute to improved health outcomes are typically unclear. 29 In this exploratory study, we showed that there may be more opportunity to improve what gets talked about rather than how T A B L E 5 Variation in patient participation across decision topic it is discussed. Specifically, we found opportunities to expand oncology discussions through the pursuit of more explicit, shared decisions about topics in which uncertainty is great and in which patient context, preference and belief system are especially relevant. To the extent these topics affect patient desires, efforts and abilities for self-care, it is not difficult to imagine pathways through which their inclusion could affect health outcomes.
Still, further study is needed to evaluate the meaning and significance of our findings, ideally in other settings, with larger samples, and through different methods. Inductive, qualitative study could be especially helpful, particularly in attempting to understand what patients and clinicians are counting as "decisions." Practically, it is important to understand the feasibility and opportunity costs of changing the discussion in oncology practice. Indeed, efforts to fully align agendas, elicit contextual information and increase patient and clinician involvement require time. Integrating these activities into the care delivery system (without sacrificing discussion that oncology clinicians must prioritize to ensure that care is meeting safety and efficacy targets) will be challenging. This may justify the development and trying of conversation aids that encourage discussion and decisions related to more diverse "cancer care"-related topics.