Characteristics of uncertainty in advanced pediatric cancer conversations

Abstract

Objectives:

To describe the initiation, response, and content of communication about uncertainty in advanced pediatric cancer.

Methods:

Qualitative analysis of 35 audio-recorded outpatient consultations between physicians and families of children whose cancer recently progressed. We defined uncertainty as “future-oriented lack of clarity in which answers are unknown to all participants involved in the conversation.”

Results:

Conversations contained a median of 14 (interquartile range 8.5–19) uncertainty statements related to 6 topics: appropriateness of treatments, acute toxicities and morbidities, prognosis or response to treatment, diagnostic uncertainty, logistical uncertainty, and long-term toxicities. Physicians initiated 63 % of statements (303/489), parents initiated 33 % (165/489), and children initiated 2% (10/489). We identified 14 unique responses to uncertainty: 11 responses provided space for discussion, 3 responses reduced space. Physicians most commonly responded by providing additional information (38 %, 361/ 947). Parents most often responded with continuing statements, such as “um” or “yeah” (50 %, 313/622). Children seldom responded (<1%, 12/1697).

Conclusion:

Physicians initiated most uncertainty discussions, and their responses often provided space for further discussion. Children were seldom involved in these conversations.

Practice implications:

Clinicians should consider maintaining open conversations about uncertainties in advanced pediatric cancer, and consider engaging children in these discussions.

1. Introduction

Uncertainty is inherent in advanced pediatric cancer. From diagnosis through survivorship or end of life, families and clinicians must navigate serial uncertainties. As noted by Cohen, “Uncertainty varies in degree of magnitude, intensity, and saliency—from the overarching, existential issues of life and death to the inconsequential contingencies and probabilities that are the substance of everyday life.” [1] Clinicians can help guide families through these uncertainties with myriad responses. Yet, few studies have examined these responses to uncertainty in pediatric oncology. [2] These limited studies show that parents of children with certain cancers feel great distress during and after completion of therapy, and anticipatory guidance might decrease this distress. [1,3] Additionally, illness uncertainty is associated with decreased quality of life for children with cancer [4], and intolerance of uncertainty is associated with heightened worry and anxiety [5]. Parents have also identified managing uncertainty as a core function of communication [6,7]. In adult oncology, uncertainty has been associated with “interferences in daily life, fear, emotional distress, reduced resourcefulness, a sense of losing control, and diminished quality of life.” [8] In these adult patients, poor cognitive ability, less education, and limited social support have been associated with greater uncertainty [8]. Depending on developmental status, children can have limited cognitive capacity and low levels of education. Thus, children might be especially at risk for the adverse effects of persistent uncertainty.

Mishel developed a theory of uncertainty in illness in the 1980s that has been foundational to our understanding of how patients experience and adapt to illness uncertainty. [9,10] However, this model focuses on an individual’s internal processing of uncertainty. Few studies have explored how clinicians and families work together to navigate shared uncertainties during clinical encounters. By understanding the context and content of these uncertainty conversations, clinicians might be able to better support these families through the multitude of uncertainties inherent in pediatric cancer.

In the following study, we aimed to characterize the content, initiation, and responses to shared uncertainty in pediatric oncology. We performed directed content analysis of 35 audio-recorded clinical conversations between clinicians and families of children with progressive or relapsed cancer.

2. Methods

We report this study following Consolidated Criteria for Reporting Qualitative Research guidelines [11] (Appendix 1).

2.1. Participants and recruitment

In this paper, we present a secondary analysis of 35 audio-recorded conversations from a study examining clinician-family communication at the Dana-Farber Cancer Institute and Boston Children’s Hospital (Boston, MA) from February 2010 to October 2013. [12,13] Families were eligible if the consenting parent spoke English, had a child younger than 18 years of age with advanced cancer that had recurred or was refractory to first-line therapy, and if they had a planned meeting with the child’s oncologist to discuss this diagnosis. Parents of children with a first relapse of acute lymphoblastic leukemia were ineligible in the larger study because of higher possibility of cure. All parents provided written informed consent, and all children older than 10 years provided assent. All parents and children were aware of probable disease relapse or progression prior to these recorded conversations. Parents were invited to participate in person or by phone. A total of 44 (56 %) of 79 eligible parents agreed to participate. Thirty-five conversations were audio-recorded; the remaining conversations were not captured due to timing and availability of staff. We did not return results or transcripts to participants.

2.2. Data collection

Conversations between clinicians and families were audio-recorded and professionally transcribed. Study staff members came before and after encounters to place and remove audio-recorders; however, they were not present during the conversations. Study authors did not participate in the placement of the audio-recoding equipment or in the conversations with families. The institutional review board of the Dana-Farber Cancer Institute approved this study.

2.3. Data analysis

Codebook development occurred in three phases using Dedoose qualitative software. In consultation with all authors, two authors (BAS and ABF) performed all coding. One coder (BAS) was a clinical fellow and the other (ABF) was a graduate student in healthcare ethics. Both coders had past training and experience in qualitative analysis. Prior to coding, these coders engaged in a reflexivity exercise to acknowledge and address personal biases.

First, these authors defined and identified instances of shared uncertainty. For the purposes of this analysis, we defined shared uncertainty as “a future-oriented lack of clarity in which the answers are unknown to all participants involved in the conversation.” This definition included situations in which probabilities were known (e.g. 20 % chance of side effect) but it was uncertain whether the child would experience the event. We defined an uncertainty statement as “a factual statement or question that initiates a discussion of the shared uncertainty.” An “uncertainty statement” initiated an “instance of uncertainty”. This “instance of uncertainty” included the initiating statement and all sequential responses to the same initiating “uncertainty statement.” To determine whether all participants in the conversation were uncertain about a topic, we relied on analysis of the surrounding content of the discussion. We excluded excerpts in which one participant was able to provide a certain answer.

Next, two authors (BAS and ABF) utilized directed content analysis [14,15] to develop a codebook for the content of uncertainty conversations. These authors first read through the transcripts to familiarize themselves with the content, then created categories of content through iterative cycles of refining codes and definitions.(Table 1) After the codebook was agreed upon by all authors, these same two authors consensus coded all transcripts for content of uncertainty. Independent coding of the interviews and comparison of coding assured both credibility and dependability of the coding process.

Table 1.

Coding Definitions for Uncertainty Statements.

Definition of uncertainty statement: Initiating statement of a future-oriented lack of clarity where the answers are unknown to all participants in current conversation.
CODE	DEFINITION
Appropriateness of Treatments	Uncertainty about whether one qualifies for a given treatment from a medical/health standpoint, or whether a given treatment is appropriate and necessary for a given patient’s situation. This includes clinical trials and standard (or nonstandard) chemotherapy if the eligibility is determined by clinical status, but not when determined by availability of trial slots. This type of uncertainty is present before a given treatment is started. This excludes uncertainty about the outcomes of the given/proposed treatment.
Acute Toxicities, Morbidities, and Psychosocial Challenges	Uncertainty about toxicities and morbidities related to the cancer and/or treatment in the near future. This includes physical and psychosocial difficulties while being treated for cancer, but it excludes long-term toxicities or morbidities the patient will experience. This type of uncertainty occurs once a given treatment has been initiated. This uncertainty can be related to the effects of cancer on the patient or family. To apply this code, the clinician or parent should use the language of uncertainty (e.g. “might”, “should”, “often”, etc.) This code does not apply to declarative statements of toxicities, etc. (e.g. “This treatment will cause/does cause/results in …”)
Prognosis or Response of Disease to Treatment	Future directed uncertainty about the child’s likelihood of responding to therapy, surviving, or being cured of cancer. This also includes uncertainty about future progression or spread of cancer.
Diagnostic Uncertainty	Uncertainty about final diagnosis, staging, or other characteristics of current disease. This uncertainty can be resolved once testing has resulted. This uncertainty must be related to present or future state. It does not relate to parents describing past uncertainties that are now resolved.
Logistical Uncertainty	Uncertainty related to scheduling or planning procedures, treatments, or other aspects of cancer care. This includes eligibility for clinical trials when enrollment is solely determined by availability of trial slots, not clinical status of the patient. This also includes questions about anticipated length of stay in the hospital.
Logistical Uncertainty	Uncertainty related to scheduling or planning procedures, treatments, or other aspects of cancer care. This includes eligibility for clinical trials when enrollment is solely determined by availability of trial slots, not clinical status of the patient.
Long-term Toxicities	Future-directed uncertainty about the toxicities and morbidities that the child will experience in the future, after treatments have already been completed. This excludes acute toxicities from treatments.

In a third phase, two authors (BAS and ABF) again used directed content analysis to develop a codebook to identify the initiation of and response to uncertainty. We used the Verona Coding Definitions of Emotional Sequences (VR-CoDES) as an a priori theoretical model for codebook development. VR-CoDES is a consensus-based coding framework for identifying emotional cues and concerns in audio-recorded clinical encounters. [16,17] VR-CoDES categorizes responses to emotion as either providing further space for exploration and conversation, or reducing space. We modified definitions and codes to apply to uncertainty rather than emotion. We employed a semantic approach by coding only topics that were explicitly articulated, and we adopted a realist ontology. Employing a similar strategy, we identified the individual who initiated the discussion of uncertainty and whether each response to this uncertainty provided or reduced space for further discussion (Table 2). We defined “providing space” as a response that continues or expands discussion of the uncertainty. We defined “reducing space” as a response that limits or redirects away from further discussion of the uncertainty. After the codebook was agreed upon by all authors, these same two authors consensus coded each transcript and reconciled any differences through discussion between the coders. A single uncertainty statement often generated a series of responses, thus the total number of uncertainty initiations does not equal the total number of responses.

Table 2.

Coding Definitions for Responses to Uncertainty.

Definition of response to uncertainty: A response to uncertainty is present when the respondent directly addresses content of original uncertainty statement. A single uncertainty statement can lead to a series of responses across multiple conversational turns.
Type of response	Definition
Reducing space: Statements that discourage further discussion or exploration of a particular uncertainty
Deferring	Actively deferring the conversation about a particular uncertainty to another time (e.g. “I’d like to talk about this with you later.”)
Redirecting	Responding to uncertainty by shifting focus to a new topic that is not linked to original uncertainty statement (e.g. [In response to a prognostic uncertainty statement] “We really have three treatment options.”)
Shutting down	Ending discussion of a particular uncertainty by explicitly suggesting against discussing the uncertainty (e.g. “Let’s not talk about that.”)
Providing space: Statements that encourage or facilitate further discussion or exploration of a particular uncertainty
Acknowledging	Directly addressing the unknown without providing explicit estimates (e.g. “We just don’t know what will happen.”)
Agreeing or accepting	Making a statement or phrase that indicates agreement or acceptance of conversational counterpart’s statement (e.g. “Yes, I agree” or “I see what you mean.”)
Clarifying	Asking questions to elicit further information about the particular uncertainty (e.g. “When will we know if this is working?”)
Contingency planning	Advising an action or behavior that depends on the outcome of an uncertain future event (e.g. “If this treatment doesn’t work, we can talk about clinical trials” or “If he develops nausea, we will give him antiemetics.”)
Guessing or Estimating	Offering a probability statement or likelihood of a particular uncertain outcome (e.g. “He will most likely lose his hair” or “His counts might be down for a week or two after the treatment” or “About 40 % of children will get this side effect.”) Use of the word “hope” with an estimate or probability would also qualify as guessing or estimating.
Hoping	Making a statement about a desired outcome without offering a probability or statistical estimate or without explicitly guessing at outcomes. (e.g. “We are certainly hoping this will work.”)
Ensuring monitoring or further evaluation	Requesting or offering additional tests, examinations, and/or monitoring in response to a particular uncertainty (e.g. “We will check his labs weekly so we know if his counts drop.”)
Offering continuing statement	Offering any word or phrase that signals listening without contributing content or context. These statements allow the discussion to continue. (e.g. “mm-hm,” “yeah,” “OK”)
Paraphrasing	Reiterating or paraphrasing counterpart’s prior statements from conversation, potentially demonstrating close listening or understanding. (e.g. “Yeah. We don’t know how he will do.”)
Providing additional information	Providing general information as it pertains to the uncertainty. This could include technical information (e.g. explaining a treatment protocol), personal information (e.g. past experiences with a specific treatment), or anticipatory guidance (e.g. what to expect in the future)
Providing options or recommendations	Advising specific actions to be taken now. This is different from contingency planning, which indicates actions to take if a future event occurs. (e.g. “I would recommend X”)

3. Results

3.1. Participant characteristics

We analyzed 35 clinical encounters that involved 35 consenting parents, 30 children, along with 34 physicians, 6 nurses, 5 nurse practitioners, and 1 psychologist. One encounter included an interpreter because one parent did not speak fluent English. The majority of participating parents were white (77 %) and female (83 %). The majority of participating children were male (69 %) with a median age of 10.3 years. The most common diagnoses were solid tumors (54 %, 19/35) and brain tumors (40 %, 14/35), with a minority of children diagnosed with hematologic malignancy (6 %, 2/35). Children were present for most of the clinical encounters (26/35 present for full clinical encounter, 4/35 present for part of clinical encounter). The median age of the 5 children who were not present was 7.3 years, ranging from 3.3–12.6 years, compared to a median age of 10.3 years and a range of ages from 2.5–17.5 years for children who were present. For 60 % of conversations (21/35), the clinical team was the primary oncology team for the family. For the remaining 40 % (14/35), the family was pursuing a second opinion. (Table 3)

Table 3.

Participant and Encounter Characteristics.

Participant Characteristics
Child characteristics
Age at enrollment (median, range)	10.3 years (2.5–17.5 years)
Female gender, n (%)	11 (31 %)
Diagnosis
Solid tumor, n (%)	19 (54 %)
Brain tumor, n (%)	14 (40 %)
Hematologic malignancy, n (%)	2 (6%)
Parent characteristics
Female gender	29 (83 %)
White, non-Hispanic	27 (77 %)
Encounter Characteristics
Length of Conversations
Mean (Standard Deviation)	71 min (29 min)
Range	16 to 123 minutes
Primary or second opinion
Primary oncology care, n (%)	21 (60 %)
Second opinion, n (%)	14 (40 %)
Number of participants in encounter
3	8 (23)
4	12 (34)
5+	15 (43)
Number of encounters that included participants
Family
Parent/Step-parent	35 (100)
Child	30 (86)
Grandparent(s)	4 (11)
Sibling	1 (3)
Clinicians
Physician	34 (97)
Nurse	6 (17)
Nurse Practitioner	5 (14)
Psychologist	2 (6)
Interpreter	1 (3)

3.2. Initiation of uncertainty conversations

All conversations contained at least 1 uncertainty statement with a median of 14 (interquartile range 8.5–19). Many uncertainty statements generated a series of responses. (See Fig. 1 for an example.) In total, we identified 489 total uncertainty statements in all audio-recorded conversations. Physicians initiated 63 % of statements (303/489), parents initiated 33 % (165/489), children initiated 2 % (10/489), and others initiated 2 % (11/489). (Fig. 2) Some initiations of uncertainty conversations took the form of a question: “So surgery isn’t something we’d have to look at right now? It wouldn’t be a dumb decision not to do a surgery.” [Father] Other initiations took the form of a factual statement: “So they have a newer [treatment], which is a [modified treatment], that is a phase II trial that might be open around the time he would be through all of this, which might be something we might think about.” [Physician]

Fig. 1.

Example of Coded Excerpt.

Discussions of uncertainty often occurred across multiple conversational turns between parents and clinicians.

Fig. 2.

Initiations and Responses to Uncertainty.

Panel A: In addition to the initiations indicated in this figure, we also identified 5 nurse initiations and 6 initiations from other family members. These were not included in the figure to ensure clarity of the physician-parent-child comparison. Panel B: We also identified 15 nurse responses (14 provided space, 1 reduced space), 2 psychologist responses (both provided space), and 11 responses from other family members (10 providing space, 2 reducing space). These were not included in the figure to ensure clarity of the physician-parent-child comparison.

3.3. Topics of uncertainty

Our analysis of uncertainty statements reached thematic saturation after identifying 6 different topics. (Table 4) The most common topic was appropriateness of treatments (27 %, 133/489). The questions of appropriateness tended to focus on whether the child would qualify for a treatment study, whether the treatment would be safe for the child, or whether the treatment was the best available option. For example, “I think, you know, it’s your child and I don’t know if you can ever have a comfort level with this but you have to be comfortable with your decisions and since there is not a right answer, it is hard.” [Physician] The next most common content involved acute toxicities, morbidities, and psychological challenges related to cancer or its treatment (26 %, 129/489) and prognosis or response to treatment (20 %, 98/489). Discussions of toxicities mostly focused on estimating the probability that the patient would experience the side effect: “Mother: Any thinness? Paleness? Physician: Well the paleness may come if he becomes anemic as a result of the chemotherapy. For the most part, it should be okay.” Some discussions of prognosis were direct: “That’s what I kind of figured anyways. But she kept asking me so I figured I’d ask anyway. The other one, and you probably can’t answer it clearly anyway, is what’s the prognosis?” [Mother] Other discussions of prognosis were vague or indirect: “So it is not something that we can do again and again and again. We can do it again and try to get the symptoms improved, improve his gait and things like that. That would be the goal.” [Physician] Diagnostic uncertainty (10 %, 51/489), logistical uncertainty (10 %, 50/489), and long-term toxicities (6 %, 28/489) were discussed less frequently.

Table 4.

Characteristics of Uncertainty Statements. Each excerpt represents a separate statement that initiated a discussion of uncertainty (i.e. instance of uncertainty). Only responses to uncertainty (Table 5) were categorized as providing or reducing space.

	Total Number of Codes (%)	Example
Type of Uncertainty (n = 489)
Appropriateness of Treatments	133 (27 %)	“So, um, you know, I am sure you guys have been dealing with this for a couple of years now, so I am sure you know the approach to relapsed [cancer]. It is not uniform. There is not, um, a best … necessarily a best way to go.” (Physician) “What is his best thing to do? Is it to head to 5 days of chemo or leave it alone? If he leaves it alone and waits, does it get worse or should he do the chemo and hopefully it stalls it and we find out ok maybe like you said that there is a chemo we can hit [child] with …” (Father)
Acute Toxicities, Morbidities, and Psychosocial Challenges	129 (26 %)	“Is he going to feel a little tired? Probably, but hopefully he will bounce back pretty quickly.” (Physician) “This won’t make my hair fall out, will it?” (Child) *Note - Child was receiving new medication where it was uncertain if hair would fall out. “And, um, during the infusion is there anything that you really have to worry about?” (Father)
Prognosis or Response of Disease to Treatment	98 (20 %)	“It is possible that you could get rid of everything. It is just that we have never seen anybody get rid of everything and get rid of everything for good.” (Physician) “Are you guys sure this is going to work?” (Child) “So here is my question and [Doctor]. Would say like why do you ask me these things? Um, can this cure him or it is just to tame it?” (Mother)
Diagnostic Uncertainty	51 (10 %)	“And the problem is a lot of things can look like tumor or tumor can look like normal brain.” (Physician) “They said you were in remission and now you have signs of activity. I said hold on a minute. Back in August, we weren’t technically in remission and now we are sort of relapsing. Um, so, you know, there has just been kind of a lack of clarity about exactly where she is.” (Father)
Logistical Uncertainty	50 (10 %)	“Just while we are talking about it, the one thing with the, um, [research study], which is what we are talking about, um, it is not open yet … Um, and there is just some administrative things with them. We think we will get it worked out in the next two weeks I am hoping.” (Physician) “So treatments of any kind, whatever the treatment may be, um, she’s not going to have to be monitored or anything like that, like she’s not going to have to stay down here, is she?” (Father)
Long-Term Toxicities	28 (5%)	“For reproductive risks in terms of whether he can have children, these medicines don’t typically affect the ability to have children, but if he is on them for long period of time, we won’t know.” (Physician) “The compromising effects of all the treatment cumulatively, is that like a long-term thing? In a year from now, would he still have issues or would he at least be able to come back to his baseline?” (Father)

3.4. Responses to uncertainty

In total, we identified 1697 responses to these uncertainty statements. Physicians provided the most responses to uncertainty statements (56 %, 947/1697), followed by parents (38 %, 622/1697). Children seldom responded (<1 %, 10/1697). Nurses (<1 %, 15/1697), psychologists (<1 %, 2/1697), and other family members (<1 %, 11/1697) were minimally involved in discussions of uncertainty. Notably, the total number of responses to uncertainty (1697) was not equal to the total number of initiating uncertainty statements (489) because clinicians and parents often responded with a string of responses to a single uncertainty statement. On average, every uncertainty statement had 3–4 responses. (Fig. 2)

3.4.1. Providing space

We identified 1579/1697 (93 %) responses that provided space for further exploration. We further categorized these responses into 11 distinct categories. (Table 5) Physicians most commonly responded by providing additional information (38 %, 361/947). For example, “Um, but, at the same time, that response or, um, tumor shrinkage is what we would want. Um, but we would be just as happy with the tumor staying the same because there have been those changes in the tumor and it looks like it had gotten a little bit bigger.” [Physician] Acknowledging (13 %, 119/947), estimating (12 %, 114/947), and providing options (11 %, 107/947) were the next most common responses for physicians. Each of the remaining physician responses represented 2%–6% of all physician responses. (Fig. 3)

Table 5.

Characteristics of Responses to Uncertainty.

Type of Response	Example (Response is underlined)
Providing Space
Acknowledging	Mother (through interpreter): Is it guaranteed that with the chemotherapy it will shrink? Physician: Oh is it guaranteed? No, unfortunately there is no guarantee. I can’t promise that this chemotherapy will shrink, ah, the tumor.
Agreeing or Accepting	Physician: I think that um … I think it would be worth more talking to our stem cell group and getting their opinion. I mean my opinion as a [cancer] person is that she is in such good shape and she has stem cells and has never gotten a transplant that it might be worth trying. Mother: I think that might be the most logical step.
Clarifying	Physician: Nothing would be terrific. Um, a little bit left, just fine. Probably, and this is where you want to be a little careful. I want to be careful because I don’t want to set the wrong expectation, but you would like it to be at the bare minimum at least 50 % gone and closer to 90 % gone. But it kind of depends on how quickly, you know, you reach a certain point where maybe you say you have about 3 quarters of it gone and you look with the next PET scan after another one or two of these and it is about the same, then we have to make a decision about whether that is good enough or whether we have to switch to whatever comes after this. Mother: When you say good enough what does that mean?
Contingency Planning	Physician: And he could do it, yeah. You know, there is some chemotherapy that can cause, um, like neuropathic pain, you know, like the Vincristine that he got way back when and sometimes that can make his walk funny, um, and it is not a big deal if we needed to treat. Um, the Irinotecan and Temodar are not really known for that but it is kind of what it seems like to me. Um, yeah, but if it keeps on getting worse, bring him back in there and they can do scans sooner rather than later to see what’s up.
Ensuring Monitoring, Follow-up, or Further Evaluation	Physician: “So, the [medication], um, it can affect your liver, kidney function, but we will be monitoring that closely.”
Guessing or Estimating	Father: Now could this put him back in remission? Physician: So it could put him back in remission. Yeah. It could definitely put him back in remission. Um, I think it is an overall 40% response rate or so and probably 10–15% of those are complete remissions.
Hoping	Mother: Even if he gets this [treatment], I shouldn’t be thinking, ok he is going to get cured. This is the last one and that is it? Physician: Yeah. Unfortunately, no. But the good part is we do expect him to feel well and he looks like he is feeling just fine. So we hope to continue that really. You know, this treatment has relatively few side effects. So we hope that he is going to continue to keep feeling this good. And we hope even more that his [imaging] will look better.
Offering Continuing Statement	Physician: So just in terms of time, I estimate it will be about 3 weeks of treatment. So really about half of the time. Um, he may lose some hair. Mother: Mmm-hmm.
Paraphrasing or Reiterating	Nurse: I would say like in regards to the experience, our goal is to keep it away as long as we can and what we end up doing is we end up switching and switching and switching. It is not like we are done. We are switching and switching and if we stop, most likely it will come back. So it can be years, you know, five or seven years but the hope is and we talked about that, maybe in that 5 years, this really has worked and this works ….You know, and that is our goal and that is why you kind of change things and that is why I think that [name] brings up a good point that we are doing something that he hasn’t seen. I think that really would be a positive thing. We are always kind of thinking you want to hit it with something different. Father: Something similar to what he responded to, but different.
Providing Additional Information	Physician: Okay, so we did talk about the immune system and the bone marrow effects on the body? Mother: Yes, we did talk about that. Which is why we didn’t do etoposide earlier even though [Doctor] really wanted us to.
Providing Options and/or Recommendations	Mother: If itworks say exceptionally well, let’s say it almost gets rid of everything.Would that mean he would go off treatment for a few years? Physician: Well, it is always a hard thing to know. We would probably recommend some other kind of treatment after he finishes it. And the question is what? And it depends a little bit on what is left over to be honest with you. But if we could get rid of all of it or most of all of it, we might recommend something kind of low-dose or oral. Something more of a maintenance kind of thing.
Reducing Space
Deferring	Mother: How come he can’t just do the [chemotherapy] and then do the radiation? Physician: He could and I want to get to that in just a moment because we are going to talk about radiation.
Redirecting	Physician: So it is too early for the sarcoma to arise in the radiation field, right? So my guess is that this is a rhabdomyosarcoma, but just that it looks a little bit different. But for this trial, he has to have a rhabdomyosarcoma. Mother: Ok so to be on this trial, you are not supposed to be on this therapy for like two weeks before the trial starts, correct?
Shutting Down	Father: That is part of the lack of understanding that we have about this that I think has raised our level of concern is while everyone doesn’t seem to be doing a good job talking to each other, is this thing just exploding inside of her? Physician: Yeah. I really don’t think that it is. We will take a quick look at her but she looks great and that marrow wasn’t done very long ago. It was only done two weeks ago, the 7th I think. Is that right?

Fig. 3.

Categories of Responses to Uncertainty.

Deferring, redirecting, and shutting down were all responses that reduced space for further discussion. The remainder of responses were all coded as providing space for further discussion.

Parental responses followed a different pattern. Half of parental responses were continuing statements (50 %, 313/622), which we defined as “offering any word or phrase that signals listening without contributing content or context. These statements allow the discussion to continue. (e.g. “mm-hm,” “yeah,” “OK”).” The next most common parental responses involved providing additional information (17 %, 108/622) and asking clarifying questions (15 %, 91/622). The information parents provided tended to relate to observations about their child or explanations of their beliefs. For example, “She had a port” [Father], and “He had severe stomach cramps.” [Mother] Each of the remaining parent responses represented 0 %–7 % of responses. (Fig. 3)

Responses from the child were sparse, including clarifying questions (50 %, 5/10) and providing information (33 %, 3/10). In response to a discussion of acute toxicities, for example, one child asked the following clarifying question: “Will it be any worse than it was the last time?” [Child]

3.4.2. Reducing space

We identified 118/1697 (7 %) responses that reduced space for further exploration. We further categorized these responses into 3 distinct categories. (Table 5) When physicians reduced space, they most commonly deferred discussion to a later time or another clinician (46 %, 26/57). For example, “So I don’t know if you guys want to come back tomorrow. I am here all week if you want to maybe come back and talk more in-depth.” [Physician] (Fig. 3) The next most common response was redirecting the conversation away from the uncertainty (39 %, 22/57). For example, after initially discussing an uncertain prognosis, a physician quickly transitioned to discussion of toxicities: “Um, we also don’t know for sure yet what all of the side-effects are.” Rarely, physicians shut down conversations of uncertainty (15 %, 9/57). For example, one physician paused a conversation of uncertainty by saying, “Um, maybe I shouldn’t be saying all these things with you [child] here, huh? I am solidifying your opposition. [Laughs]” [Physician]

When parents reduced space, they most commonly redirected conversations away from the uncertainty (74 %, 23/31). Sometimes, this redirection was accomplished with humor. For example, one parent responded to the physician’s discussion of potential toxicities by stating, “I feel like I have suffered from that [laughs]. You can check mine.” [Mother] Rarely, parents shut down the conversation in response to uncertainty (23 %, 7/31). For example, one father interjected while a physician was discussing potential toxicities: “Well, at this point in time, we’re assuming, everything we’re seeing, there’s a chance of really bad stuff happening, we know that.” [Father]

Children only demonstrated 2 responses that reduced space. One response redirected away from uncertainty and 1 response shut down the conversation.

4. Discussion and conclusion

4.1. Discussion

Parents and physicians in this study had open discussions about a multitude of uncertainties falling into 6 content categories. Additionally, we identified 14 unique responses to uncertainty: 11 that provided space for further discussion and 3 that reduced space. Physicians in this study did not seem hesitant to acknowledge and address uncertainties. In fact, physicians initiated most discussions about uncertainty, and their responses to uncertainty seldom reduced space for further discussion. While there is not yet empirical evidence that families benefit from these responses that provide space, many families in prior work have reported the importance of receiving difficult information about prognosis [18,19] and late effects [20,21]. Additionally, findings from adult medicine have found that physician uncertainty expressions were associated with greater patient satisfaction. [22] Similarly, many families have indicated the importance of knowing what to expect and developing plans to address uncertainty [7]. We observed all of these behaviors in our analysis.

Our finding of predominantly open discussions of uncertainty differs from findings in adult medicine. Primary care physicians in one study were reluctant to discuss uncertainty because of concerns about negative consequences and ambiguity aversion. [23] In a study of breast cancer patients, physicians only disclosed prognosis uncertainty to half of their patients [24]. This difference in openness between pediatric and adult physicians could result from different training or selection practices in each field. This difference could also relate to the family-centered model of care in pediatrics that promotes openness and engagement with the family. Lastly, it is possible that this coding framework was not sensitive to responses that reduced space, or families might have avoided verbalizing more concerning uncertainties. Future studies should aim to better delineate these spoken and unspoken uncertainties.

Notably, physicians most often responded to uncertainty by providing information. Providing information is a central role for pediatricians in communication and often meets the needs and desires of families. [6,7] However, other responses to uncertainty might also support families during periods of uncertainty, especially when the uncertainty will be longstanding. This emphasis on information might indicate different views of diagnostic closure for physicians versus parents. In past work, Cohen found that physicians viewed diagnostic certainty as providing closure to uncertainty. Parents viewed diagnostic certainty as providing an answer to one question, but “it does not provide any sense of closure nor suggest how they should proceed.” [1] Perhaps clinicians in our study viewed information as providing closure to uncertainty, yet parents experienced persistent or new uncertainties even after provision of information. Although information provision is paramount, we suspect that many families in uncertain situations will benefit from receiving acknowledgment and support in addition to information.

The uncertainties we identified in this study ranged widely in emotional burden and potential duration. For example, uncertainty about timing of chemotherapy initiation likely triggered fewer emotions than uncertainty about survival or serious toxicities. Furthermore, some uncertainties were time-limited, whereas others would persist for years, decades, or an entire lifetime. For example, many diagnostic uncertainties would be resolved upon further imaging or biopsy, whereas worries about secondary cancers or other limitations later in life could persist for the remainder of the child’s life.

Although we did not identify emotional exploration among the array of responses to uncertainty, this approach could potentially be helpful in dealing with the emotional burden of uncertainty. For example, Kimbell et al. demonstrated the effect of pervasive uncertainty on coping strategies of adults with advanced liver disease. The authors recommended that clinicians address this uncertainly directly in a constructive manner to support coping and ameliorate patient distress. In another study of men with prostate cancer, a watchful waiting intervention that addressed uncertainty was associated with decreased confusion, improvement in quality of life, and viewing “their lives in a new light.” [25] Similar approaches in pediatric oncology might support family’s uncertainty while also acknowledging the emotional burden of adapting and planning for uncertainty. Snaman et al. have suggested using “what if” language to address the worries associated with uncertainty [26]. This language, for example, might normalize the distress associated with uncertainties: “For some families, talking about these worries can be helpful. Would it be helpful for you? Would you like to talk about this now?” [26] Future work should further explore the psychological effects of these different responses to uncertainty.

Our study was not designed to explore the phenomenological experience of uncertainty, but our findings suggest that parents experienced an open and engaging environment for talking about their worries and fears. In prior work, parents have described how strong relationships can support their management of uncertainty. [7] Actions that support these relationships (e.g. demonstrating care and concern, demonstrating competence and reliability, advocating, and maintaining open and reassuring nonverbal communication) might similarly support parents in their ability to manage uncertainty. Creating these caring networks are especially important for parents of children with cancer, because worries about side effects, late toxicities, and relapse can linger for many years [27].

Despite this apparent open discussion between parents and physicians, children were seldom involved in these conversations. Children only initiated 2% of these conversations, and they offered less than 1% of responses to these uncertainties. In pediatric oncology, parents are dealing with a child’s life-threatening illness, making complex decisions about care, balancing their own emotions with their child’s need for support, and managing long-term care needs that create significant burdens. In this complex and stressful environment, parents and children must also manage many types of uncertainty, as we have demonstrated in this study. Some children with cancer are young and might have concrete concerns about the present or the near future. Children in adolescence, however, can develop “existential plight” (intense emotional distress, usually focused on life-and-death concerns and physical symptoms). [28–32] Older children will likely require different approaches to uncertainty than younger children. Furthermore, a limited body of literature in pediatrics has shown that illness uncertainty is associated with decreased quality of life for children with cancer [4]. However, little is known about children’s preferences or needs when addressing uncertainty. It is possible, for example, that children choose not to engage because these topics are confusing or worrisome. Future studies should aim to learn from children and adolescents with cancer to understand how to help them best address uncertainty, but we anticipate that children will benefit from further engagement in these conversations.

The literature on managing uncertainty in pediatrics is sparse outside of this current study. Our analysis contributes to an empirical understanding of how clinicians and families navigate uncertainties in the clinical setting, but many questions remain. We do not know, for example, how families personally experience the myriad uncertainties inherent in serious childhood illness. Nor do we know how parents and children communicate with each other about these unknowns, or how they manage the resultant angst and worry. Furthermore, we do not know how clinicians experience this uncertainty. Medical schools train physicians to make diagnoses and find answers. Does current medical training prepare physicians to engage with patients about persistent uncertainties? Lastly, we do not know if our findings represent the experiences of families with children at different time points in the cancer trajectory. The family of a child near diagnosis or in survivorship seems likely to experience similar types of uncertainty, but these clinical encounters might unfold in unique ways. These important questions merit dedicated study.

Our study should be interpreted in light of several limitations. First, we focused on outward expressions of uncertainty that were shared by the clinicians and family members. As such, we are unable to comment on internal uncertainties held by any participants. The Hawthorne effect might have also biased communication processes if participants communicated differently because they knew they were being recorded. Furthermore, only 56 % of parents we approached agreed to participate, which could contribute to selection bias. Additionally, we utilized audio-recordings and thus were unable to assess for nonverbal aspects of communication. Lastly, we analyzed one conversation per patient for uncertainty statements. While these initial conversations after relapse or progression contained many uncertainty statements, the responses to these uncertainties could have unfolded over multiple discussions.

5. Conclusion

Clinical conversations about advanced pediatric cancer in this study contained varied types of uncertainty. Physicians initiated the majority of uncertainty discussions, and their responses to uncertainty often provided space for further discussion. Despite being present, children were seldom included in these conversations.

6. Practice implications

Clinicians should strive to maintain open conversations about uncertainties in advanced pediatric cancer, while striving to engage children more fully in these discussions. Additionally, clinicians should consider acknowledging and exploring the emotional toll of uncertainty.