Treatment Recommendations in Oncology Visits: Implications for Patient Agency and Physician Authority

Abstract

Although oncology is a major site for clinician-patient treatment negotiation requiring a careful balance of potentially competing viewpoints, little is known about how clinicians promote their treatment recommendations to patients and what the manner of promotion tells us about the oncologist-patient relationship. Utilizing an already-established schema of coding treatment recommendation, I draw on 61 treatment recommendations to examine treatment decision-making in oncology. This paper investigates how physicians balance asserting their authority while at the same time attending to patient agency and involvement in decision-making. Taking this one step further, this paper explores how physicians negotiate decision-making with patients given that they occupy a liminal state between obligations to policy imperatives and commitments to their professional knowledge and technical expertise. How do they do this, and what accounts for this? To answer these questions, this paper analyzes the ways in which physicians present treatment recommendations and the treatment contexts in which they are made.

A primary function of clinical consultations with ill patients is to identify an appropriate treatment. Physicians hold most of the cards in this situation: they have the knowledge to diagnose and to understand what the diagnostic implications are for treatment, and they possess the legal and cultural authority to prescribe (Stivers, 2002; Timmermans & Oh, 2010). Yet, a shift in patient-provider dynamics since the 1960s has provided patients with increasing power. The rise of patient consumerism and informed consent has re-oriented the medical profession toward increased patient involvement (Jones, 2017). The increasing availability of health information on the internet has also facilitated increased patient involvement in their own health care (Boyer & Lutfey, 2010). Some have pointed out that a consequence of this is the “democratization” of health information through its availability to both laypeople and health-care professionals (Anspach, 2011). Moreover, patients are the practical gatekeepers of all actual treatment implementation insofar as they must ingest pills, show up for chemotherapy, etc.

This yields an interdependent relationship between physician and patient, making treatment decision-making an exceptionally rich communicative environment for examining the delicate balance of patient and physician viewpoints in how decisions get made. This is particularly the case in oncology because treatments are complex and range from innocuous to toxic, from symptomatic to life-saving. A patient’s refusal of treatment could lead to increased morbidity and mortality, but it also can represent an autonomous choice within their rights. I will show that physicians approach delivering treatment recommendations in nuanced ways that work to secure patient acceptance - prioritizing medical authority in some cases and patient agency in others. I discuss what insight this provides us into how physicians orient to different recommendations depending on where patients are in the process of their treatment.

Literature review

Early advocates of the shared decision-making (SDM) paradigm idealize the decision-making process as collaborative: physicians and patients are involved in the treatment decision through a two-way information exchange, communication of preferences, and joint investment in the treatment decision (Légaré & Thompson-Leduc, 2014). The physician is charged with providing information regarding all available treatment options, risks and benefits, and impacts on patients’ well-being. Conversely, patients must offer their preferences, values, beliefs, and burden tolerances (Charles, Gafni, & Whelan, 1999). While physicians are required to mobilize their expertise in SDM through communicating their treatment preferences, the model represents an erosion of physician authority insofar as it runs counter to the model of paternalistic decision-making through promoting patient autonomy and choice (Brown, Butow, Butt, Moore, & Tattersall, 2004).

This model represents one of many structural changes to the medical institution that strips physicians of some of their once-total control of the medical institution (Starr, 1982). Of course, physicians continue to possess particular medical expertise and technical knowledge specific to their profession (Timmermans & Mauck, 2005), which they bring to the table when implementing SDM. Moreover, they too have preferences and beliefs about what is best for their patients, sometimes with significantly greater experience to draw on about the possible efficacy and side effects of a treatment. Given this, physicians arguably walk a tightrope between their obligations to facilitate patient choice and the mobilization of their medical expertise in treatment decision-making.

Yet how physicians navigate patient involvement in decision-making varies. In recent years, conversation analytic scholars have described the process of treatment decision-making in adult primary care (Koenig, 2011), pediatrics (Stivers, 2007), and neurology (Toerien, Shaw, & Reuber, 2013). Further, how patients account for resistance to (and rejections of) treatment recommendations has been empirically studied (Lindström & Weatherall, 2015), as have physician accounts for treatment recommendations (Angell & Bolden, 2015). Other conversation analytic scholars have explored features of treatment recommendation design. Kushida and Yamakawa (2015) have observed that different proposal types are used in different contexts of medical decision-making. In their data, while the proposal type “Let’s” or “How about” is used to make a treatment decision, the declarative evaluation proposal type, (e.g., “It may be better to switch to another drug”) is used to cautiously propose treatment.

Besides proposals, other actions are used to produce treatment recommendations. Investigating treatment recommendations in U.S. and U.K. primary care, Stivers et al. (2017) established a typology of five treatment recommendation actions drawn from their data: pronouncements, suggestions, proposals, offers, and assertions. The authors argue that each has features which communicate different degrees of physician authority and patient agency in the decision (see Table 1).

Table 1.

Treatment recommendation actions from Stivers et al. (2017).

Coding dimension	Options (if applicable)	Explanation	Example (if applicable)
Social action
	Pronouncement	Physician asserts recommendation as instigator and decision-maker and presents as already determined	“I’ll start you on X”
	Suggestion	Physician recommends as instigator but treats patient as decision-maker and medication as optional	“You could try X”
	Proposal	Physician recommends as instigator but decision-making is treated as shared by doctor and patient.Proposals highlight the recommendation as speculative	“Let’s try X and see how that goes”
	Offer	Physician treats patient as having instigated recommendation and as the decision-maker, thus treating medication as having been occasioned	“Would you like me to give you X”
	Assertion	Physician’s asserts a generalization about a treatment’s benefit implying a recommendation but not proffering an overt directive	“X is good for this”

Pronouncements (e.g., “I will start you on X”) assert full deontic authority (Stevanovic & Peräkylä, 2012) as they frame the treatment recommendation as already decided by the physician and as if the patient has no choice in the matter. Suggestions (e.g., “You could try X”) allow for the patient to ultimately decide whether or not s/he is going to act on the recommendation while still establishing that the physician has agency in the decision insofar as it has originated with him/her. Proposals (“Let’s try X and see”) are like suggestions insofar as they originate with the physician and do not direct the patient (as pronouncements do). But, proposals uniquely treat the decision as shared between physician and patient. Offers (e.g., “Would you like me to give you X?”) exert low physician authority over the decision, treating the patient as instigator of the recommendation and frame the recommendation as something the physician is prepared to do for the patient as a response to a patient issue. Finally, assertions (e.g., “X is good for this”) are statements of fact about a treatment’s benefit that imply a recommendation without personalizing the recommendation for the patient. When applying this schema to U.S and U.K. primary care, Stivers et al. ultimately found that U.S. physicians rely more heavily directive actions like pronouncements, while U.K. physicians are more likely to use inclusive recommendations like proposals and offers. This suggests a greater willingness to exercise authority when recommending treatment and a reciprocal de-emphasis on patient agency in the U.S., with the inverse occurring in the U.K. In specialty contexts in the U.K., this trend was even stronger with a greater emphasis on forms that amplify patient agency in psychiatry (Thompson and McCabe (2017) and neurology (Toerien, 2017).

I wondered whether analyzing the occurrence of these treatment recommendation actions in a different treatment environment - oncology - would add an additional layer of explanatory power to the schema established by Stivers et al. (2017). After all, oncology is a unique treatment environment. A cancer diagnosis is generally perceived as life-threatening. Soon after diagnosis, patients need to make major time-sensitive decisions with a care provider with whom they do not yet have a relationship (Hillen, Onderwater, van Zwieten, de Haes, & Smets, 2012). Further, cancer patients have a disease which could progress slowly or rapidly, and no one can be fully certain whether it will respond to standard treatments. Because of the onerous nature of the treatments and the essential uncertainty of the disease itself (Garofalo, Choppala, Hamann, & Gjerde, 2009), patients require constant and close monitoring by their oncologist. This necessitates an especially close partnership between oncologist and patient.

In this area of highly consequential decision-making, SDM may be especially warranted based not only on its relationship to better health outcomes (Corser, Holmes-Rovner, Lein, & Gossain, 2007) but also on the fact that patients prefer it (Charles, Whelan, Gafni, Reyno, & Redko, 1998). SDM involves a negotiation of decision-making between physicians and their patients. Although any time physicians direct a patient through recommending treatment they exercise some deontic authority, the question is, to what degree do they assert this authority in oncology?

This study seeks to explain how physicians balance mobilizing their medical expertise and authority while preserving patent agency in oncologic decision-making. I first identify the different ways that physicians present treatments to patients. Then I ask, to what degree do physicians involve patients and under what circumstances do they mobilize their professional authority? I subsequently examine whether recommendation actions are associated with treatment decision types, specifically examining the context of the decision and whether a patient’s stage in the treatment process matters in how physicians communicate with patients. This allows us to assess whether there are meaningful patterns in the way physicians orient to patient involvement across different types of decisions and how physicians reflect different norms for patient rights and responsibilities in the cancer treatment process.

Method

All participants provided written informed consent and all study procedures were Institutional Review Board (IRB)-approved. The data come from 90 video-recorded oncology visits in the western U.S. collected between 2014 and 2017. Each visit was transcribed according to standard conversation analytic (CA) conventions (see Hepburn & Bolden, 2012), and data analysis was guided by CA practices (see Sidnell & Stivers, 2012). Recordings were examined for initial treatment recommendations. These included initial recommendations for single treatments only. Recommendations for both cancer and pre-cancer diagnoses were included. Recommendations for surgical interventions, prescription drugs including chemotherapy and radiation, and over-the-counter medications were included. Recommendations for testing were excluded as were treatments outside of the cancer diagnosis. Recommendations in response to a patient request for a treatment (in “second position”), or redone recommendations for the same treatment, were excluded. Using these parameters and applying the coding schema for treatment recommendation actions established by Stivers et al. (see Table 1), 61 instances of treatment recommendations met the inclusion criteria. Examples of each action type from these oncology data and comparable primary care action types are detailed in Table 2.

Table 2.

Treatment recommendation action examples in oncology.

TR action	TR action in primary carea	TR action in oncology
Pronouncement	“I’ll start you on iron tablets”	“You’ll get both infusions today”
Suggestion	“You know you could try Claritin for that”	“We strongly recommend you consider using the birth control pill for prevention”
Proposal	“Why don’t we put you on the plain Allegra”	“1 still think we need to do chemotherapy”
Offer	“Would you like me to give you a stronger pain killer”	“We can give you some lidocaine jelly and see if that helps”
Assertion	“This medication that will take care of [it] is called Actifed”	“There’s still utility in taking the ovaries out of people with BRCA2”

^aExamples taken from Stivers et al. (2017)

Following the coding of actions in oncology, I observed that there was a relatively higher frequency of proposals and suggestions in visits where the patients were new to treatment and a relatively higher frequency of offers when patients had complaints that were secondary to the central treatment. Thus, I subsequently coded for the treatment stage in which recommendations were produced, which I refer to as treatment context, as an additional analytical variable. I identified three primary treatment recommendation contexts in these data: new, mid-course, and ancillary (see Table 3).

Table 3.

Treatment recommendation contexts in oncology.

Treatment context	Definition	Example
New	Treatment that has not yet begun for a cancer-related issue	Postsurgery to discuss next steps in treatment; initial visits where mass has been found; prophylaxis for a genetic predisposition to cancer
Mid-course	Recommendation for an intervention for an already begun oncology treatment course	Already begun chemotherapy, radiation, testosterone suppression, or steroid treatment
Ancillary	Treatment for some medical issue ancillary to the primary cancer- related diagnosis	Treatment for a side effect of nausea with chemotherapy; treatment for pain near incision area after surgery

Results

The context of treatment recommendations in oncology

The question I ask is whether most treatment recommendations in oncology orient to collaboration between physician and patient through the use of proposals - which, relative to the other four treatment recommendation actions proposed by Stivers et al. (2017) - best embodies an implementation of SDM, or whether the pattern is more typical of U.S. primary care and relies primarily on pronouncements. A plurality of recommendations (39%) were made via pronouncements, 30% were via offers, 21% via proposals, 7% via suggestions, and just 3% via assertions. Notably, although patient agentive actions predominate in these data when combined (61%), it is striking and somewhat counterintuitive how frequent pronouncements are in the oncology context. Recall that pronouncements prioritize physician authority, delivered as though patient responses were unnecessary. Significantly, this high rate of reliance on pronouncements in the sample runs counter to the literature’s recommendations, which advocates for a collaboration between physician and patient. Is this purely an artifact of U.S. physicians’ proclivity to rely on high authority recommendations? Another possibility is that treatment recommendation actions are shaped by the type of treatment being recommended. Current SDM literature does not differentiate between different types of recommendations by stage in the treatment process, yet doing so underscores nuances of decision-making, patient involvement, physician authority, and the relationship to treatment context.

Indeed, I found a clear relationship between the treatment recommendation action and the context of the treatment recommendation action. Significantly, pronouncements are not made in the new treatment context. They exist only in the mid-course treatment context, where they are predominate, and ancillary treatment contexts, where they are used when physicians themselves bring up the complaint being treated (see Table 4). This observed relationship in these data speaks to how physicians orient to patient agency in different types of decisions based on the type of recommendation action they use, which I will now elaborate.

Table 4.

Treatment recommendation actions by context.

	New	Mid-course	Ancillary
Action
Pronouncements	0% (n = 0)	75% (n = 6)	51% (n = 18)
Suggestions	22% (n = 4)	0% (n = 0)	0% (n = 0)
Proposals	61% (n = 11)	12.5% (n = 1)	3% (n = 1)
Assertions	11% (n = 2)	0% (n = 0)	0% (n = 0)
Offers	6% (n = 1)	12.5% (n = 1)	46% (n = 16)
Total (n)	100% (n = 18)	100% (n = 8)	100% (n = 35)

Total single treatment recommendations: n = 61.

Recommendations for new treatments

In these data, when a patient has not yet begun treatment for a cancer-related issue, the clinician’s recommendation is for a new cancer treatment. Of the total treatment recommendations, 30% (n = 18) involve recommendations for new treatments. In every case, a proposal action (n = 11), suggestion action (n = 4), assertion action (n = 2), or offer action (n = 1) are used by the physician. All of these actions prioritize patient agency relative to the strikingly absent pronouncement actions.

In extract (1) a recommendation to begin treatment is made during a postsurgical consultation. The patient is a 38-year-old female who has recently undergone surgery to remove her uterus, fallopian tubes, and ovaries. The results from the surgery came back as cancerous, so the patient is seeing the oncologist to discuss next treatment steps. Prior to line 01, the physician discusses with the patient that she presented her pathology results (positive for cancer) to the others in her oncology group. In line 01, the physician begins to report the results of the meeting:

The physician states in lines 01–05 the treatment recommendation that the team came up with, which she also frames as her treatment recommendation: “the: team agreed that the best treatment would be a combination of chemotherapy and radiation at the same time and then followed with chemotherapy.” Note that by invoking the team’s consensus on treatment, the physician conveys that this is not only her recommendation for a future course of treatment, but that of her colleagues as well. This treatment recommendation action is a proposal: she states that the “best treatment would be a combination of chemotherapy and radiation”; “would be” treats the implementation of the recommendation as subject to a final decision, inviting the patient to both take part in the process and agree. The physician’s endorsement of the recommendation is clear both in that this is “best” and the fact that this was a consensus (“the team agreed”). This is a “new” context because the patient has had surgery but has not yet pursued treatment for her diagnosis.

In contrast to the previous case, in extract (2), a young female patient is seeing the gynecologic oncologist to discuss a recently diagnosed BRCA1 mutation. She has previously undergone a double mastectomy to reduce her breast cancer risk and is now discussing how to minimize the likelihood of developing gynecologic cancers, though she does not currently have cancer. The physician enters into her treatment recommendation (lines 07–08) after discussing the standard screening protocols (lines 01–04):

In lines 04–05, the physician begins to build the treatment recommendation as a way for the patient to prevent ovarian cancer and provides statistical evidence bolstering this claim. The core recommendation occurs in the next turn-constructional unit, (TCU) (lines 06–08): “so: if you’re not trying to get pregnant we strongly recommend that you consider just using the pi:ll.”. This course of treatment, to take the birth control pill for a reduction in ovarian cancer risk, is framed as a new course of action through a suggestion action. “We strongly recommend that you consider just using the pi:ll” includes explicit physician endorsement “we strongly recommend” and indicates that the ultimate decision is being left up to the patient “that you consider just using the pi:ll”.

In new treatment contexts, although physicians may convey strong endorsement of a recommendation, what is most critical for this analysis is that patient involvement and agency are prioritized through the actions. Across the 18 new treatment recommendations in the sample, no pronouncements are used and only action types which maximize patient agency are used in this context. The majority (n = 15) are proposals and suggestions. Rarely are offers and assertions used - which orient to patient involvement as even higher than in suggestions or proposals.

One account for this reliance on patient agentive recommendation actions in the new context is that physicians are particularly oriented to the need to secure patient acceptance of treatment. Patients could potentially resist this due to the burdensome nature of many of these treatments. Beginning cancer treatment following diagnosis is time-sensitive. At the same time, physicians have to contend with patient fears of treatments such as chemotherapy and radiation which have known and serious side effects (Passik, Kirsh, Rosenfeld, McDonald, & Theobald, 2001), making patient persuasion, and securing buy-in, especially pertinent. In sum, the preponderance of suggestions and proposals provides strong evidence for physician orientation to the patient’s ultimate decision-making capacities and a physician’s willingness to collaborate with the patient even though a physician’s tilt toward one course of action is clear.

Recommendations for mid-course treatments

In the sample analyzed, 13% of instances (n = 8) involved a recommendation for some modification to an already begun oncology treatment course. These treatments included, for instance, chemotherapy treatments, testosterone suppression therapy, and steroid therapy to treat cancer. These treatments are thus mostly the same sorts of treatments that can be initiated in new treatment contexts but are considered “mid-course” because clinicians are adjusting them in some way after a patient has agreed to the initial treatment. Notably, nearly all (n = 6) of the treatment interventions in these contexts involve a pronouncement treatment recommendation action – the action type that is entirely absent in the new treatment context– even though the treatment types are effectively the same (e.g., chemotherapy, radiation). This finding represents a shift as it indicates more of a parallel to the findings in primary care (see Stivers et al., 2017) where a majority of pronouncements were used to recommend treatment. Yet, the medical issues present in this context are of an entirely different nature than those seen in primary care settings.

In turning to the next encounter, extract (3), we see just this. Most recommendations in this context are responsive to a drug inefficacy problem, like pain or bleeding from continued tumor growth. However, in this case, the physician is responding to the worrisome side effects of the drug treating this patient’s cancer instead of the drug’s inefficacy or cancer growth. The patient, a 48-year-old male, is dealing with recurrent metastatic prostate cancer and is being treated for it with both chemotherapy and steroid treatment. Earlier in the visit, the patient had complained of nerve pain, which his physician associated with the strong steroid (Ozurdex) he had been taking. In extract (3a), we see the patient complaining of an “electrical charge” sensation in lines 02–03:

The physician’s subsequent treatment recommendation in extract (3b) - to wean the patient off the Ozurdex and replace it with a weaker steroid - is motivated by the patient’s complaint of this uncomfortable sensation. The treatment recommendation is made in lines 01–02:

The physician uses a pronouncement action for the treatment recommendation (lines 01–02): “as you’re coming off of this we’re gonna- (.) we’re gonna switch you over to prednisone?” This recommendation, for a change to steroid treatments from the Ozurdex to prednisone, is categorized as a pronouncement action because the physician frames the recommendation as if it has already been decided and as though the patient has no choice in the decision. Note that even when the dosage of the prednisone is discussed just after in lines 07–10, the physician continues to use a pronouncement action in determining the dosage: “we’re gonna do:. prednisone five milligram tablets” (lines 07–08) and provides an account for doing so, that the steroids give him unpleasant side effects: “I’m gonna keep you a little on the low side David cuz I think this stuff messes ya up a little bit” (lines 08–10). The physician is therefore oriented to both the change to prednisone and its dosage as already determined and as if the patient has no choice in the matter.

The next encounter, in extract (4), offers another instance of a pronouncement. The patient, a middle-aged female patient, is currently being treated for metastatic ovarian cancer. She has complained of pain in her abdomen which her oncologist believes is due to continued growth of the tumor. Before this visit, she has undergone two cycles of chemotherapy which has been shown to be ineffective in treating her tumor. In lines 15–16, the physician pronounces the treatment recommendation:

The physician takes the patient’s complaint of pain (line 10–11) and uses it as leverage to connect the pain to the inefficacy of the chemotherapy (lines 12–13). The patient confirms that the pain is still not improved since chemotherapy has been started (line 14). In the next turn, the physician provides a recommendation to change the treatment (lines 15–16) in the context of this continued complaint of pain. This mid-course treatment recommendation, from the existing chemotherapy drug to a new chemotherapy drug (which we later learn is Doxil), is made via a pronouncement: “this time we’re gonna try: a-a new dru:g that you have not seen before = hopefully will not make you so nauseated” (lines 15–17). Due to the existing drug’s inefficacy, a change to the treatment mid-course is pronounced, framing the decision as already determined and treating this patient as if she has no say in the matter. Recall that in addition to these key features, pronouncements also display high physician agency and patient acceptance is unsought.

In this context, physicians rarely use actions other than pronouncements, such as proposals, to recommend treatment. This result suggests that once patients have signed on to being treated for this cancer, and have forged a relationship with their physician, physicians treat their level of agency as of lower priority than in the new treatment context. Instead, they privilege their own deontic authority in the mid-course context. These data suggest that, once patients sign onto treatment, physicians orient to patient agency as having been somewhat transferred to them.

A deviant case

The claim thus far is that physicians prioritize patient agency at the outset of treatment in an environment where they treat

patients as needing to get on board. Contrastively, once a patient has begun treatment and shown a willingness to follow recommendations, physicians typically rely on pronouncements - which treat patient agency as less important and instead prioritize physician authority. The importance of this lies in the fact that physicians in these data do not give patients an explicit opportunity to sign-off on treatment adjustments. Without explicitly seeking buy-in at each treatment adjustment period, there is little opportunity given to patients to consider alternative therapies, stopping treatment, or shifting to palliative or supportive care.

Evidence for this interpretation can be reinforced through looking at a deviant case, extract (5). Here a physician relies on a proposal for a mid-course treatment recommendation, countering our identified pattern. However, the account for this works to strengthen the analysis because it shows that mid-course treatment recommendations can be understood as a point where the physician’s relationship with the patient is being established.

This patient has been undergoing maintenance therapy for ovarian cancer, but her cancer has recently grown. Prior to this, she was being treated by another oncologist and it is unclear why she has switched to this oncologist - whom she saw in a prior visit before today’s visit. In the following extract, after discussing getting the records from the patient’s previous doctor, the physician enters into her treatment recommendation (lines 07–09):

The treatment recommendation, “so I think carboplatinum will be fine for you like we talked last time” (lines 07–08) is categorized as a proposal because it is framed as an idea originating from the oncologist, “I think that”, but non-definitevely and is looking for approval from the patient. It is also framed as something they had discussed before but, because it is being retopicalized now, we can deduce that a course had not been formally decided in prior to this. This physician is seeking the patient’s explicit buy-in to the recommendation, which does not fit with most of the recommendations in this context - pronouncements. Recommendations mid-course are typically handed down to the patient and presuppose patient agreement. In contrast, this recommendation invites patient agreement.

When we look at what is new about this treatment encounter, it is not the context in which the recommendation is being made, rather it is the relationship between the physician and patient. This supports the idea that when a new physician-patient relationship begins, as in the new treatment context, the physician orients to the need to get the patient on board with treatment. While this encounter seems to depart from what is usually done in this context, it in fact helps clarify what is salient about the different contexts: It is not just about whether the treatment is familiar or unfamiliar, it is also about the degree of familiarity in the physician-patient relationship. Here, because the physician and patient are relatively unfamiliar to each other, the physician orients to the need to get this patient’s buy-in before changing the drug treating her cancer.

Recommendations for ancillary treatments

Ancillary treatment recommendations are the third type of recommendation observed in these data. These treatments are remedies to medical complaints which are ancillary to the primary diagnosis. For instance, the complaint could be a side effect of chemotherapy, like nausea or neuropathy, or a complaint associated with surgery. In these data, 57% of recommendations were of this type (n = 35). The salient aspect of an ancillary treatment context is that the treatment is not explicitly treating the existence of the cancer but treating a side effect of some sort.

Ancillary complaints can come at the initiative of the patient or at the initiative of the physician and this matters for the way the problem is addressed. Thus, these ancillary treatment contexts were further divided into categories separating whether the recommendation related to the ancillary treatment was made at the physician’s initiative or in response to a complaint presented by the patient. When these were divided, 43% of ancillary treatment cases (n = 15) were made at the physician’s initiative, and 57% were made at the patient’s initiative (n = 20). Notably, all of those treatment recommendations given preemptively by the physician were done via the pronouncement recommendation action. Conversely, those given in response to patient complaints were primarily offers (n=16), with few pronouncements (n=3) and one proposal. I suggest that the action type again indexes physicians’ priorities: when physicians initiate the consideration of an ancillary problem, they treat their medical authority as primary, whereas when patients initiate consideration of an ancillary issue, patient agency is underscored.

An instance of a recommendation done in the context of an ancillary complaint made at the physician’s initiative is offered in extract (6). At the outset of the visit, she registers that the patient does not look good and may need to be admitted. The patient, in the middle of chemotherapy for ovarian cancer, has presented with dizziness, fever, and bloating. The topicalization of the concern begins in lines 01–02:

The pronouncement for an ancillary treatment (IV fluids) comes at line 51, “we’ll getchya some IV fluids”. The recommendation is ancillary because it is a treatment meant to deal with a concern secondary to her cancer treatment, a concern that includes looking a little “dry” and being “warm to the touch” (lines 9 and 11). After the patient identifies another symptom in line 46, “I am a little dizzy,” the physician aligns with the patient in her turn at line 47. Following this, the physician tells the patient that she is “pretty worried” about her (line 49). Notably here the physician takes the initiative to treat the ancillary problem by topicalizing the concern herself. The problem is also framed as an urgent one in need of attention. In this context, this physician asserts full medical authority and hands down the treatment recommendation via a pronouncement rather than collaborate with the patient.

Another instance of a pronouncement recommendation action produced at the physician’s initiative is shown in extract (7). The patient is to begin her first chemotherapy treatment during this visit for recurrent ovarian cancer, and the physician is reviewing the process of the chemotherapy infusion, for instance, that she will be given some medicines to help her rest during the infusion. She also reviews which chemotherapy drugs are to be administered during the infusion. She then asks the patient whether she filled her prescription for the anti-nausea pills, and then enters into a treatment recommendation for the patient to take those pills:

Once the physician has confirmed that the patient did fill the prescription already given to her for anti-nausea pills and they are readily available to her (lines 01–05), she does a positive assessment (lines 06–07). This case is categorized as a treatment recommendation made at the physician’s initiative because the physician enters into her treatment recommendation to take the anti-nausea pills (lines 07–10) before any complaint or discussion about nausea has been made. Further, this case is classified as an ancillary treatment context because the anti-nausea pills are only indirectly related to the primary disease, which is ovarian cancer in this case. The anti-nausea pills are treating a side effect of chemotherapy, the primary treatment for the patient’s disease. Notably - like all other ancillary treatment types in which the physician preempts the medical complaint - the treatment recommendation is done as a pronouncement.

These pronouncements treat the recommendation as not optional. It therefore follows that the preponderance of pronouncements in this ancillary treatment context suggest that physicians are strongly privileging their medical authority and experience, have a clear stance on what the patient ought to do, and therefore demonstrate more deontic authority in making these types of recommendations. This could be motivated by the need to prevent risk. In the first extract, the physician wanted to prevent further risk to the patient’s health as she appeared clinically weak, while in the second, the physician wanted to prevent the risk of nausea because of the chemotherapy drug.

The next instance in extract (8) is also an instance of a recommendation made in an ancillary treatment context; however, this time the treatment recommendation action is an offer. This patient has recently had surgery to remove her uterus, tubes, and ovaries after cancer was found in her cervix. She is back in the clinic to discuss the pathology results with the physician. Before this happens, however, she presents with some pain around the incision area:

The patient first establishes in lines 01–02 that she is feeling discomfort around the incision area in her initial

presentation of the problem. Next, the physician examines her (line 05) and counters, stating that the incision area looks “perfectly normal” (line 08) which gives the patient a “no problem” response to her initial complaint of pain. This sets the stage for the offer action of the treatment recommendation for a numbing topical cream (lines 32–33): “We can give you some lidocaine jelly and see if that helps with it”, which comes after more discussion of the nature of the patient’s pain. Note that the physician has not registered any concern for the issue and is unable to explain why the pain is there (lines 31–32), thus she frames the treatment recommendation as something that is being constructed in response to an issue presented - this time, at the patient’s initiative.

Another instance of an offer in the ancillary treatment context is shown in (9). This elderly female patient has a tumor in her lower abdomen for which she is being treated with chemotherapy. Her primary complaint is of rectal bleeding during bowel movements, which we learn is the tumor pushing on her rectum:

After the physician explains and subsequently normalizes the rectal bleeding (lines 01–05), she begins a treatment recommendation that is framed as helping with this complaint of bleeding (lines 07–08): “.hhh um we can give you medicines to help with the:. s-> you know to< make them SOFTER and see if that helps us.” This recommendation, given in the form of “I can give you X to help”, is clearly made an offer responsive to the patient’s complaint. The key is, however, that the physician does not endorse the solution. Instead, she supports the solution to the problem in response to a complaint made at the patient’s initiative.

Thus, in offers, even though the treatment recommendation is verbally launched by the physician, it is constructed as a remedy in service to needs originating from the patient. Notably, in the majority of ancillary context cases where the complaint was made at the patient’s initiative, the physician recommended the treatment in the form of an offer, or something s/he would be doing in service to the patient. Physicians do not orient to sharing in the decision, rather they orient to their role as gatekeeper to the solution of the complaint.

Discussion

This paper examines how the actions that physicians deploy to recommend treatment reveal their orientations to the role of patient agency in treatment decisions. We have observed that physicians treat patient agency as central to initial decisions about cancer treatment and ancillary treatment for patient concerns, but physicians orient to patient agency as relatively less important for mid-course adjustments of cancer treatment or for ancillary treatments that they initiate. By examining how physicians orient to balancing physician authority and patient agency, we see how they walk a tightrope between obligations to policy imperatives to collaborate with patients and commitments to advocating for what they think is best based on their professional knowledge and technical expertise.

Initially, in considering other work done on treatment recommendation actions, the oncology data in this sample patterns most similarly to the primary care data analyzed by Stivers et al. (2017) where pronouncements were used most often. This is interesting given that oncology is considered secondary care, yet the findings here are, on their face, dissimilar to treatment decision secondary care contexts more generally (Thompson & McCabe, 2017; Toerien, 2017).

When an additional layer of analysis was undertaken to consider the context of the decision (e.g., new, mid-course, or ancillary), a salient finding emerged that accounts for this difference. When stage in treatment process was considered and the data were divided into three recommendation contexts (new, mid-course, and ancillary treatments), it was clear that oncologists differentiated between recommendation actions depending on the context. In this way, physicians demonstrated different

orientations to patient participation and thus patient agency depending on the type of decision. Specifically, physicians considered where patients were in the process of their treatment and what the type of recommendation was (i.e., whether it was related to the cancer itself or the result of a side effect). The results suggest that physicians reflect different norms for patients’ rights and responsibilities to be involved in decision-making depending on these variables. In the data, oncologists deployed more agency and were more authoritative in recommendations for changes to treatment and less so when making recommendations for new treatment. Significantly, the deviant case shown, a proposal made mid-course, points to evidence that the relationship between physician and patient - and not just the treatment stage - is also at play in analyzing orientations to patient involvement across contexts. This case is evidence for the relationship between physician and patient superseding the context in which the treatment is given.

This analysis provides meaningful insight into how physicians view their relationships with patients by finding that physicians are quite nuanced in how they present treatment recommendations. For instance, physicians typically provide initial recommendations for radiation, chemotherapy, or surgery as proposals, inviting patients to align with this plan. Yet, when explaining to patients that existing treatment needs to be adjusted - including major changes - physicians typically use pronouncements that tell patients what will be happening as in “I’m going to switch you to X.” This suggests that once clinicians get initial buy-in from patients, they have carte blanche for subsequent medical decisions.

This also means that mid-course in treatment there is less interactional room for patients to put their own treatment preferences on the table, preferences which could include implementing alternative therapies, stopping the current treatment, or considering palliation. Instead, this analysis suggests that the system has the potential to set patients up, even with advanced stage cancers, for prolonged treatment. From a practice standpoint, this offers preliminary evidence supporting the problem of an underuse of palliative and hospice care in oncology (Levit, Balogh, Nass, & Ganz, 2013). This analysis points to the way oncologists see their role with established patients - more authoritative - as a potential source for the U.S. health-care system’s overtreatment problem in cancer care.

This paper also builds on an interpretation of reduced resistance correlated with the use of pronouncements originally proposed by Stivers et al. (2017). The authors speculate that less resistance after pronouncements is either conditioned by the pronouncements themselves or that pronouncements are used in contexts where resistance is expected to be minimal. This analysis suggests that in environments where resistance was more likely, such as at the outset for cancer treatment, physicians were not using pronouncements. In environments where minimal resistance was anticipated, like when patients have already begun treatment, physicians did use pronouncements.

The lack of pronouncements in the new treatment context could also be explained in terms of securing trust in the treatment regimen and subsequently adherence to it. Does handing down a treatment recommendation ex-cathedra by pronouncing it truly secure patient buy-in like a proposal or suggestion would? Proposals and suggestions involve patients explicitly in the final decision, and thus involvement in decision-making is linked to buy-in. If patients explicitly get on board with these oftentimes burdensome treatments, they could be more likely to adhere to the treatment regimen then those who are simply told what is going to be done. While proposals could allow for more resistance to the treatment, as was found in primary care, suggestions do not see similar rates of resistance (see Stivers et al., 2017). Perhaps, physicians are willing to open up the possibility for a treatment negotiation if it means ultimately securing buy-in up front.

Conclusion

In the realm of empirical investigations of physician-patient interaction, this paper uniquely focuses on physician orientation to patient involvement in decisions about cancer treatment and how this reflects medical norms in the preservation of patient rights in a domain of complex decision-making. This analysis also suggests that treatment context in decision-making is important. More importantly, these data highlight that the consideration of the different types of decisions in the treatment process matters significantly in understanding how decisions are made. At a fundamental level, physicians’ behaviors appear to be shaped by their understandings of the physician-patient relationship at different stages of the treatment process. This work builds on foundational studies of how treatment recommendations are actually produced across different clinical contexts and implications for facilitating SDM (see Barnes, 2017; Bergen et al., 2017; Stivers et al., 2017; Thompson & McCabe, 2017; Toerien, 2017).

This analysis of oncologic decision-making contributes to the next line of future research on the production of treatment recommendations. In analyzing oncology, we have preliminary grounds for understanding the additional impact of treatment context in how decisions get made between physician and patient. Furthermore, this work offers a foundation for future studies that measure the relationship between patient/physician sociodemographic variables and the way treatment recommendations get delivered. Subsequent investigations might look more closely at variables like patient and physician gender, patient socioeconomic status, and patient race/ethnicity and their causal link to physicians’ orientations to patient involvement in decision-making. This future work could help us to better address existing disparities in care delivery across different marginalized groups by pointing to gaps in patient involvement and ways to improve the clinician-patient relationship in cancer care.