The Evolution of Regret: Decision Making for Parents of Children with Cancer

Abstract

Purpose:

Parents of children with cancer make treatment decisions in highly emotional states while feeling overwhelmed with information. In previous work, 1 in 6 parents demonstrated heightened decisional regret regarding treatment at diagnosis. However, it is unclear how regret evolves over time. We aimed to determine whether parents of children with cancer experience decisional regret over time, and to identify parental characteristics and clinician behaviors associated with longitudinal regret.

Methods:

Prospective, questionnaire-based cohort study of parents of children with cancer at two academic pediatric hospitals. Parents reported decisional regret at diagnosis, 4 months and 12 months.

Results:

At baseline, 13% of parents (21/158) reported heightened regret, 11% (17/158) at 4 months (p=.43, McNemar’s test relative to baseline) and 11% (16/158) at 12 months (p=.35 relative to baseline and p=.84 relative to 4 months). In multivariable analysis using generalized estimating equations adjusted for timepoint of survey completion, heightened regret was associated with non-white race/ethnicity (OR 11.57, 95% CI 3.53 to 41.05, p<.0001) and high anxiety (OR 2.01, 95% CI 1.04 to 3.90, p=.04). Parents with high peace of mind (OR 0.24, 95% CI 0.09 to 0.62, p=.003) and those reporting high quality information (OR 0.22, 95% CI 0.07 to 0.69, p=0.01) had lower odds of heightened regret. We found no association between heightened regret and time point of survey administration.

Conclusions:

A small, significant proportion of parents experience heightened regret throughout the first year of their child’s cancer treatment; non-white parents are at higher risk. Effective communication may protect against regret.

Introduction

Making informed medical decisions on behalf of their children with cancer is central to many parents’ perceptions of being a “good parent.”[6, 8] By participating in informed decision making, parents can more fully advocate for their children in situations when they largely feel powerless.[31] However, parents often report deficiencies in information and understanding about potential toxicities, late effects,[23] prognosis,[29] and participation in early- and late-phase clinical trials.[3, 12] Compounding these informational deficiencies, such decisions largely occur around the time of diagnosis or relapse, when parents and children can feel overwhelmed.[13] For example, parents have described the time after diagnosis as a blur, feeling “numb,” “paralyzed with fear,” and “in shock.”[11] Making decisions in this emotional state can lead to challenges in decision-making.

Perhaps as a result, approximately 1 in 6 parents demonstrated heightened decisional regret within 12 weeks of making initial cancer treatment decisions for their children in previous work.[17] With regret comes self-blame, a highly negative emotion and key feature of depression. Parents of minority racial/ethnic status in this initial analysis had disproportionately higher odds of reporting heightened decisional regret, whereas parents who held an ideal decision-making role, received high-quality information, or trusted the physician completely were less likely to experience this regret near the time of diagnosis.[17] In the adult literature, physician empathy and fulfilling preferred roles have also been associated with decreased decisional regret,[21] and lower satisfaction with information has been associated with heightened regret.[26] However, these studies have primarily focused on cross-sectional data proximal to the treatment decisions. As such, it is unclear how decisional regret evolves over time, or what factors might modify the risk of longitudinal regret. Some of this regret and self-blame might be allayed with directed interventions. To develop these interventions, however, investigators should understand how regret evolves over time.

In order to evaluate regret in greater depth, we sought to determine whether decisional regret persists over time, which subgroups of parents might be at a higher risk of longitudinal regret, and what clinician behaviors might protect against this regret. In this study, we evaluated a longitudinal cohort of parents of children with cancer treated at one of two pediatric oncology centers, for which baseline data on parental regret were previously published.[17] To understand prevalence and predictors of regret, we administered questionnaires focused on communication and decision-making experiences, including regret, near the time of initial treatment decisions, then 4 months and 12 months after diagnosis.

Methods

We surveyed parents and oncologists of children with cancer at Dana-Farber Cancer Institute and Boston Children’s Hospital (Massachusetts) and Children’s Hospital of Philadelphia (Pennsylvania) between November 2008 and April 2014, as previously described.[17] We asked the parent with primary responsibility for decision making to participate. If both parents equally shared decision-making roles, they chose which parent participated. Eligibility criteria included: parent able to read English or Spanish, child 18 years or younger, first contact 1 to 6 weeks from cancer diagnosis, and child’s oncologist permitted contact. Parents were offered $10 gift cards after completing each questionnaire.

Parents completed follow-up questionnaires 4 and 12 months after diagnosis. Parents were excluded if their child had died, the parent declined further contact, or the parent participated in a separate qualitative interview. Of 565 eligible parents, 382 (68%) completed baseline questionnaires. 304 parents were eligible for 4-month questionnaires (11 children died since baseline, 46 parents declined further contact, and 21 participated in qualitative interviews), and 211 completed questionnaires (69%). 205 parents were eligible for 12-month questionnaires (6 children died), and 168 completed questionnaires (82% of those eligible, and 44% of participants at baseline). Ninety-five oncologists completed matched surveys for 361 patients (95%) at baseline. Oncologists assessed prognosis at time of diagnosis, 4 months, and 12 months. 158 parents answered decisional regret questions at all 3 time points and were included in these analyses.

Data Collection

Parent and physician questionnaires included items from previously developed surveys[16, 19] and select items from existing validated instruments. Survey development was described in detail previously.[17] Questionnaires were available in paper-and-pencil or electronic format, and in English or Spanish. See Supplemental Figure S1 for a flowchart detailing survey domains included at each time point.

The primary outcome of this analysis was decisional regret, which was assessed using the Decision Regret Scale.[2] This item asked parents to reflect on decisions about their child’s cancer treatment and respond to the following prompts: “I have made the right decisions;” “I regret the choices that were made;” “I would make the same choices if I had to do it all over again;” “The choices did my child a lot of harm;” “The decisions were wise.” Response options included “strongly agree,” “agree,” “neither agree nor disagree,” “disagree,” “strongly disagree.” Mean scores were obtained using a scale of 1 to 5 (with reverse scoring as appropriate). A score of 1 indicated the least regret and 5 indicated the most regret. All scores were decreased by 1 point and multiplied by 25, creating a scale range of 0 to 100. As previously described,[15, 26] scores of 0 were categorized as “no regret;” 1 to 25 as “mild regret;” and greater than 25 as “heightened regret.” For analysis, regret outcomes were dichotomized to parents with heightened regret (score >25) versus parents with no or mild regret (score ≤25), consistent with prior work.[15, 17, 26]

Parent and Child Characteristics

At baseline, parents were asked to report their gender, age, race, ethnicity, and highest level of education. Child age and diagnosis were determined using medical records. For prognosis, oncologists were asked at each time point “How likely do you now think it is that this child will be cured of cancer,” with response categories of: “extremely likely (more than 90% chance of cure);” “very likely (75–90%);” “moderately likely (50–74%);” “somewhat likely (25–49%);” “unlikely (10–24%);” “very unlikely (less than 10%);” or “no chance of cure.”[14, 16, 35]

Parent Psychological Factors

We utilized the Hospital Anxiety and Depression Scale to evaluate anxiety, with scores ≥8 considered suggestive of the condition.[33] Peace of mind was assessed using the FACIT-Sp scale. In this scale, parents were presented with statements about their experiences (I feel peaceful; I am able to reach down deep into myself for comfort; I have trouble feeling peace of mind; I feel a sense of harmony within myself; I know that whatever happens with my child’s illness, things will be okay) and asked to report the extent to which these statements applied to them by choosing a response of extremely, very, somewhat, a little, or not at all.[18]

Communication and Decision-making

Parental decision-making preferences and actual roles were identified using previously developed survey items, with phrasing adapted to pediatric rather than adult medicine.[20] We asked parents “Which statement best describes the role you would prefer to play when decisions about treatment for your child’s cancer are made?” We then asked “Which statement best describes the role you actually played when making decisions about treatment for your child’s cancer?” We defined the “ideal role” as when the parent’s actual role exactly matched their preferred role.

We assessed perceived quality of information using a previously described scale,[10] including questions about quality of information related to treatment, prognosis, functional outcome, cause of cancer, and response to treatment. We also asked parents how they felt about “the amount of information you received about your child’s likelihood of cure.” Trust in the oncologist was assessed with a question from the Trust in Physician scale.[1, 7]

The institutional review board of Dana-Farber Cancer Institute approved this study, and informed consent was obtained from participants.

Statistical Analysis

Regret outcomes were dichotomized to parents with heightened regret versus parents with no or mild regret, consistent with prior work.[15, 17, 26] We evaluated for marginal symmetry between parent reports of regret at baseline, 4-month, and 12-month time points using McNemar’s test.

We then evaluated factors associated longitudinally with heightened parental regret. For bivariable analyses, in order to isolate effects of the independent variables, we used logistic regression with random effects models adjusted for time as a random effect. We considered independent variables at a single, baseline time point (parent race/ethnicity, education, and gender),as well as time varying covariates (high peace of mind, anxiety, physician-rated prognosis, relapsed disease status, parent fulfilled ideal decision-making role, parent holding a more active decision-making role than desired, parental report of high quality information, parental report of having the right amount of information regarding prognosis, and parents trusting the oncologist’s judgment about medical care completely). Information quality scores were summed and dichotomized at the median. The peace of mind subscale was dichotomized at a score of 4 or greater, a cutoff chosen to correspond approximately to the categories (“extremely” or “very”) that we felt generally indicated a strong sense of peace of mind, consistent with prior work.[18]

Multivariable analyses used generalized estimating equations, adjusted for the timepoint of survey assessment. We used an unstructured correlation matrix for the generalized estimating equations. We used a backward elimination technique to create a multivariable model of factors associated with heightened regret at each time point, with criteria for entry of p=0.10 and for retention, p=0.05. Models included time point, physician-rated prognosis and parent race/ethnicity regardless of the significance of their coefficients. Finally, we tested for interactions between significant covariates and time, and retained interactions with P<.05.

Upon identifying the interaction between non-white race/ethnicity and time, we subsequently assessed for differences in decisional regret by race/ethnicity at each time point using Pearson’s Chi-square test. Analyses were conducted by using SAS statistical package v9.4.

Results

Participating parents were predominantly female (83%), white (80%), and well educated. (Table 1) Most children had hematologic malignancies (52%), with the remainder having solid tumors (36%) and brain tumors (12%). Patients generally had good oncologist-reported prognosis, with 61% of patients having at least 75% chance of cure at baseline. Of the 158 parents with complete data on the decisional regret scale at all 3 time points, 3 completed the survey in Spanish; the remainder were able to complete the survey in English. Parents who completed all three questionnaires had higher educational status than those who completed the baseline questionnaire only, with 76% of those who completed all three questionnaires having completed college versus 55% of those who did not (p<.0001). Otherwise, we found no differences between the two groups with respect to parent age (p=.39), parent gender (p=.49), parent race/ethnicity (p=.47), child gender (p=.81), site (p=.82), diagnosis (p=.65), or baseline oncologist-rated prognosis (p=.23).

Table 1.

Parent and child characteristics at baseline

	N (%)
Parent age
< 30	14 (9)
30 to 39	55 (36)
40 to 49	70 (45)
50+	15 (10)
Parent gender
Female	130 (83)
Male	27 (17)
Parent race/ethnicity
White	127 (80)
Black	10 (6)
Hispanic	12 (8)
Other	9 (6)
Parent education
High school graduate or less	38 (25)
College graduate	73 (47)
Graduate/professional school	44 (28)
Parent marital status
Married/living as married	137 (88)
Other	18 (12)
Child age at diagnosis
0 to 2	40 (25)
3 to 6	37 (23)
7 to 12	42 (27)
13 to 18	39 (25)
Child gender
Male	87 (55)
Female	70 (45)
Diagnosis
Hematologic malignancy	82 (52)
Solid tumor	57 (36)
Brain tumor	19 (12)
Physician-rated prognosis (at baseline)
Extremely likely (>90% chance of cure)	34 (23)
Very likely (75–90%)	57 (38)
Moderately likely (50–74%)	38 (26)
Less than moderately likely (<50%)	20 (13)
Site
Boston	115 (73)
Philadelphia	43 (27)

n=158 with complete regret data at all 3 time points. Missing data: parent age (4); parent gender (1); education (3); marital status (3); child gender (1); prognosis (9).

Changes in Decisional Regret over Time

At time of diagnosis, 13% of parents (21/159) reported heightened regret, compared to 11% (17/159) at 4 months, and 11% (16/158) at 12 months. We found no significant difference in regret between baseline and 4 months (p=.43, McNemar’s test of symmetry), 4 months and 12 months (p=.84), or baseline and 12 months (p=.35). However, some parents had increasing regret, and others had decreasing regret over time. For example, 7% (11/158) of parents with no/mild regret at baseline reported heightened regret at 4 months. Conversely, 9% (15/159) of parents with heightened regret at baseline reported no/mild regret at 4 months. Only a minority of parents (3%, 4/158) had sustained regret throughout the entire study, but 27% (42/158) of parents experienced heightened regret at some point during this study period. (Table 2)

Table 2.

Change in Regret over Time

	4 month		Total
Baseline	No or Mild Regret	Heightened Regret
No or Mild Regret	126 (80%)	11 (7%)	137
Heightened Regret	15 (9%)	6 (4%)	21
Total	141	17	p=.43
	12 month		Total
4 month	No or Mild Regret	Heightened Regret
No or Mild Regret	129 (82%)	12 (8%)	141
Heightened Regret	13 (8%)	4 (3%)	17
Total	142	16	p=.84
	12 month		Total
Baseline	No or Mild Regret	Heightened Regret
No or Mild Regret	125 (79%)	12 (8%)	137
Heightened Regret	17 (11%)	4 (3%)	21
Total	142	16	p=.35

P value represents McNemar’s test.

Factors Associated with Longitudinal Decisional Regret

Initial bivariable analyses identified significant associations between heightened longitudinal regret and several variables of interest, as shown in Table 3. In multivariable analysis, heightened regret was significantly associated with non-white race/ethnicity (OR 11.57, 95% CI 3.53 to 41.05, p<.0001) and high anxiety (OR 2.01, 95% CI 1.04 to 3.90, p=.04). (Table 4) Parents with high peace of mind (OR 0.24, 95% CI 0.09 to 0.62, p=.003) and those who reported having received high quality information (OR 0.22, 95% CI 0.07 to 0.69, p=0.01) had lower odds of heightened regret. Neither prognosis (p=.11) nor time (p=.94 for 4 months, p=.48 for 12 months) were associated with heightened decisional regret. We also identified an interaction between non-white race/ethnicity and time, with non-white parents experiencing lower odds of regret at 4 months (OR 0.08, 95% CI 0.02 to 0.41, p=0.002) and 12 months (OR 0.11, 95% CI 0.02 to 0.60, p=0.011) relative to their experiences at baseline.

Table 3.

Factors Associated with Parent’s Longitudinal Decisional Regret in Bivariable Analyses

	OR (95% CI)	P value
Parent attributes
Non-white or Hispanic parent race/ethnicity	2.96 (1.64, 5.35)	.0003
Parent is a college graduate or higher	.55 (.30, .99)	.05
Male gender	.66 (.29, 1.51)	.32
High peace of mind	.45 (.22, .92)	.03
Anxiety	1.97 (1.09, 3.57)	.03
Disease attributes
Physician-rated prognosis >75% chance of cure	.62 (.35, 1.11)	.11
Relapsed disease	2.97 (1.11, 7.95)	.03
Decision-making and communication attributes
Parent fulfilled ideal decision-making role	.55 (.31, .98)	.04
Decisional burden - more active role than desired	3.23 (1.62, 6.45)	.001
Parent reports high quality information	.34 (.13, .86)	.02
Parent reports having the right amount of information	.27 (.15, .49)	<.0001
Parent trusts oncologist’s judgment about medical care “completely”	.30 (.17, .54)	<.0001

Evaluated using random effects models adjusted for time as a random effect.

Table 4.

Factors Associated with Parent’s Longitudinal Decisional Regret in Multivariable Analyses

	OR (95% CI)	P value
Baseline time point	Ref	-
4 month time point	1.04 (0.40 to 2.70)	.94
12 month time point	1.43 (0.53 to 3.87)	.48
Physician-rated prognosis >75% chance of cure	1.95 (0.85 to 4.43)	.11
Non-white or Hispanic parent race/ethnicity	11.57 (3.53 to 41.05)	< .0001
High peace of mind	0.24 (0.09 to 0.62)	.003
Anxiety	2.01 (1.04 to 3.90)	.04
Parent reports having received high quality information	0.22 (0.07 to 0.69)	.01
Non-white at 4 months, relative to non-white at baseline	0.08 (0.02 to 0.41)	0.002
Non-white at 12 months, relative to non-white at baseline	0.11 (0.02 to 0.60)	0.011

Evaluated using generalized estimating equations with baseline time point as reference. Models were created using the backward elimination technique, with criteria for entry of p=0.10 and p=.05 for retention. Race, prognosis, and time point were included regardless of statistical significance.

Race/Ethnicity and Decisional Regret

Given the interaction between nonwhite race and time, we evaluated heightened regret by race/ethnicity at each time point. (Table 5) At baseline, a significantly larger proportion of non-white parents reported heightened regret (10/31, 32%) compared to white parents (11/127, 9%) (p=.0005). By 4 months, the difference by race was no longer statistically significant, with 16% (5/31) of nonwhite and 9% (12/127) white parents experiencing regret (p=.28). Findings were similar at 12 months (p=.22).

Table 5.

Longitudinal Change in Decisional Regret by Race/Ethnicity

Baseline	No or Mild Regret	Heightened Regret	Total
White	116 (91%)	11 (9%)	127
Non-white	21 (68%)	10 (32%)	31
Total	137	21	p=.0005
4 month	No or Mild Regret	Heightened Regret	Total
White	115 (91%)	12 (9%)	127
Non-white	26 (84%)	5 (16%)	31
Total	141	17	p=.28
12 month	No or Mild Regret	Heightened Regret	Total
White	116 (91%)	11 (9%)	127
Non-white	26 (84%)	5 (16%)	31
Total	142	16	p=.22

P values represent Chi-square test.

Discussion

Initial treatment decisions in pediatric oncology are often made under time pressure with unmet parental information needs. Decisions made under these conditions can challenge thoughtful decision-making. We found that 13% of parents experienced heightened decisional regret at baseline, with a similar proportion over the first year after diagnosis, raising concerns about parental experiences with decision-making.

While aggregate levels of regret remained similar over time, the levels of regret for many individual parents did change. Some parents appeared to resolve feelings of regret, whereas others who were initially satisfied with decisions displayed new evidence of regret with time. Only a minority of parents reported persistent regret at each assessment during this study, but 27% of parents experienced heightened regret at least once. This finding suggests that regret can fluctuate over time, perhaps as a result of the evolving context and content of decisions being made. For example, perhaps a child’s subsequent clinical course led some parents to question their prior decisions, such as late onset toxicities or other complications. Alternatively, other parents may become disabused of prior regret if their child makes a full recovery without significant morbidity. Clinicians should be aware that decisional regret can arise at any point in the course of treatment.

Additionally, non-white parents were especially burdened with decisional regret near the time of diagnosis. Regret among minority parents could be related to lower quality communication and decision-making processes, especially if clinicians were racially discordant. Alternatively, this regret could be related to the physician’s misunderstanding or misinterpreting the parent’s cultural norms, or to parental distrust of clinicians. This racial/ethnic disparity calls for dedicated attention in future studies.

Unexpectedly, neither prognosis nor relapse status was associated with heightened decisional regret in multivariable analysis. However, receiving high-quality information was strongly associated with lower levels of regret. Although relatively few children had experienced relapse, limiting our power to definitively assess regret post-relapse, this finding suggests that the decision-making process itself is an important contributor to parental regret. As such, this finding offers a potential target for future interventions. Aiming to improve the quality of communication prior to making treatment decisions, especially focusing on improving the quality of information, could protect parents from decisional regret later in their child’s disease course. Developing a trusting relationship, responding to emotions, providing understandable information in form and amount that parents can manage, and supporting families to manage challenges at home are other targetable attributes of effective communication.[5, 27] The importance of high-quality information is reflected broadly in the pediatric[22, 24, 28, 29] and adult communication literature.[25, 34] In previous work, we found that parents largely rely on their nurses and doctors for medical information,[30] and many parents express unmet information needs throughout the first year of treatment.[29] Based on the results of our current study, these unmet information needs may be contributing to long-term decisional regret for parents of children with cancer.

Parents with high peace of mind were also less likely to report heightened decisional regret. High peace of mind could be viewed as a characteristic of parents rather than a modifiable factor. However, our previous work has shown that peace of mind at time of diagnosis is associated with high-quality communication and trust in the physician.[18] In a follow-up study of these same parents, we found that trusting the physician in the first year after diagnosis was still associated with higher peace of mind 5 years later.[32] Thus, fostering a stronger relationship between clinicians and families near the time of diagnosis might support higher peace of mind and protect against decisional regret for these parents. This finding provides another potential target for future interventions in pediatric oncology.

To date, the pediatric communication literature has largely focused on descriptive and correlational studies, with almost no focus on developing interventions to operationalize the findings from these studies. A recent systematic review of the communication literature in pediatric oncology did not identify any interventions that targeted the clinician-parent/patient interaction.[31] While this body of evidence has provided an essential, foundational understanding of communication in pediatric oncology, there is a continued need to find ways to act on these data. Our findings should provide further impetus to develop future communication interventions, especially focused on supporting the exchange of high-quality information, fostering a healing relationship, and supporting an informed, effective decision-making process that supports the individual family’s needs. Additionally, these studies should explore the role of health literacy on parental regret, as this is another actionable domain for future interventions. To achieve these outcomes, multimodal interventions that target the behaviors of clinicians and families are essential. Such interventions might include educational sessions, but should also aim to engage and empower families using question prompt lists and needs assessments to guide conversations. The use of communication technology to streamline communication with the medical team and possibly to provide reliable information directly to the patient could also support these goals. While quasi-experimental, pre/post studies can support proof of principle, multi-institutional randomized controlled trials will provide the most robust and generalizable results. As in the VOICE trial in adult oncology, multiple outcome measures will be needed to measure impact on regret, exchange of information, quality of life, and relationship with the healthcare team.[4]

Our findings, however, should be considered in light of limitations. First, the population of parents in this study was predominantly female and well educated. Additionally, patients with brain tumors were underrepresented, and this study was performed at two highly specialized academic hospitals. Thus, these results may not generalize to all clinical settings. The proportion of non-white parents was also small, which precluded more detailed analysis of individual races/ethnicities. There are certainly potential differences in the experiences of parents with different races/ethnicities, but our study was unable to examine these differences effectively. Furthermore, it is possible that cultural differences led to minority parents interpreting or responding to the questions differently than white parents. Additionally, these questions about regret do not fully reflect the complexity of feelings that parents might experience, especially related to high-stakes decisions.[9] As such, our findings provide an important starting point, but future studies should aim to more clearly elucidate the context of this regret.

The diagnosis of pediatric cancer is shocking and life altering, and cancer-related treatment decisions can have long-term implications for parents. A small but significant proportion of parents carry high levels of decisional regret over the course of the first year after their child’s diagnosis, and non-white parents are at even higher risk of feeling this regret. By focusing future studies on communication interventions to improve specific aspects of the communication and decision-making process, we may be able to address these racial/ethnic disparities while helping all parents to feel confident in their role as an advocate and “good parent” for their child.

Supplementary Material

520_2019_4933_MOESM1_ESM

Supplemental Figure S1. Flow Chart for Survey Administration