Patient-Caregiver Dyads’ Prognostic Information Preferences and Perceptions in Advanced Cancer

Abstract

Context:

Prognostic information is considered important for advanced cancer patients and primary informal caregivers to prepare for the end-of-life. Little is known about discordance in patients’ and caregivers’ prognostic information preferences and prognostic perceptions, while such discordance complicates adaptive dyadic coping, clinical interactions and care plans.

Objectives:

To investigate the extent of patient-caregiver discordance in prognostic information preferences and perceptions, and the factors associated with discordant prognostic perceptions.

Methods:

We conducted secondary analyses of a cross-sectional study (PROSPECT, 2019-2021). Advanced cancer patients (median overall survival ≤12 months) from 7 Dutch hospitals and caregivers completed structured surveys (n=412 dyads).

Results:

Seven percent of patient-caregiver dyads had discordant information preferences regarding the likelihood of cure; 24-25% had discordant information preferences regarding mortality risk (5/2/1 year). Seventeen percent of dyads had discordant perceptions of the likelihood of cure; 12-25% had discordant perceptions of mortality risk (5/2/1 year). Dyads with discordant prognostic information preferences (p<.05) and dyads in which patients reported better physical functioning (p<.01) were significantly more likely to perceive the 1-year mortality risk discordantly.

Conclusion:

Physicians should be sensitive to discordant prognostic information preferences and prognostic perceptions among patient-caregiver dyads in advanced cancer care.

Introduction

Advanced cancer threatens patients’ physical and psychological well-being, and interferes with social relationships. Primary informal caregivers (e.g., intimate partner, family member) need to adapt to patients’ disease and are confronted with an uncertain future. ^{(1, 2)} Besides facing cancer-related distress individually, patients and caregivers experience the disease as a shared stressor. ^{(3, 4)} They may cope with cancer as a unit (i.e., dyadic coping), which can ameliorate psychological and relationship functioning. ^{(1, 2)} Open communication is essential in dyadic coping, as it enables a shared understanding of treatment risks, benefits and future care goals. ^{(2, 5–8)} Apart from emotional support, caregivers assist patients in daily tasks, symptom management and medical decisions. ^{(1, 2, 5, 7, 9–11)} Caregivers’ role in patient care commonly intensifies when cancer advances. ^(6,9)

Information about prognosis (e.g., likelihood of cure, life expectancy) is important for treatment decision-making and practical preparation for the end-of-life. ^{(3, 9, 10, 12–17)} Prognostic information additionally facilitates psychological and social adjustment to death, and stimulates patient-caregiver support. ^{(12, 13, 15, 16, 18)} As individual prognostic perceptions are impacted by complex interpersonal processes (e.g., interpretation of physicians’ information), patient-caregiver dyads may not have a shared understanding of prognosis. ^{(7, 10, 19)} Generally, it is suggested that patients view prognosis more optimistically than caregivers. ^{(9–11, 14, 20, 21)} One study among dyads of older patients with advanced cancer and caregivers showed that almost one-half had discordant perceptions of patients’ survival. ⁽⁶⁾ Discordance in prognostic perceptions may induce anxiety, depression and caregiver burden; impede patient-caregiver communication, adaptive dyadic coping and relational satisfaction; and complicate clinical interactions, goal-concordant medical decision-making, and (planning of) patients’ care. ^{(2, 5–7, 10, 11, 14, 19, 22)}

Patient-caregiver discordance in prognostic perceptions possibly arises from conflicting information preferences. ⁽¹³⁾ Research suggests that caregivers are more likely to want prognostic information than patients, particularly in later disease stages. ^{(13, 15, 23, 24)} Also, dyads may base their prognostic perceptions on different data sources, especially since individuals often seek information outside of the physician. ^{(18, 25)} Furthermore, families in non-Western cultures commonly request physicians to limit prognostic communication to patients. ^{(26, 27)} Caregivers who want to hide concerns and negative information (i.e., protective buffering) or take control over patients’ care to mitigate patient burden (i.e., overprotection) may limit patients’ exposure to prognostic information and conceal discordant perceptions. ⁽²⁸⁾ Contrastingly, caregivers who deliberate patients’ emotions and information needs regarding cancer (i.e., active engagement) may facilitate concordant prognostic perceptions. ⁽²⁸⁾ Additionally, conflicting personal coping styles potentially contribute to discordance in prognostic perceptions; some people may avert confrontational stimuli like prognosis (i.e., avoidance coping) while others seek information to handle stressors (i.e., active coping). ⁽¹²⁾ Lastly, patients’ physical condition is integral in estimating prognosis, and caregivers rely on patients’ appearance. When patients exhibit few symptoms, prognostication could be more complicated, leaving room for diverging prognostic perceptions. ⁽²⁹⁾

While knowledge of patients’ and caregivers’ individual prognostic information preferences and perceptions is increasing, the agreement of prognostic information preferences among dyads is largely unstudied. The limited number of dyadic studies on prognostic perceptions paint a confined picture by isolating one aspect of prognosis (i.e., likelihood of cure or life expectancy) and focusing on the consequences of discordance ^{(10, 14, 22)}; the only paper including predictors lacked assessment of dyadic drivers. ⁽⁶⁾ To address these knowledge gaps, our study in the advanced cancer setting sought to: (1) describe the extent of patient-caregiver discordance in prognostic information preferences and (2) prognostic perceptions (i.e., likelihood of cure; 5/2/1 year mortality risk); and (3) explore the (primarily dyadic) factors associated with patient-caregiver discordance in prognostic perceptions (i.e., 1-year mortality risk).

Methods

Design

This is a secondary analysis of a cross-sectional survey study on prognostic information preferences and prognostic perceptions among patients with advanced cancer, caregivers and physicians (PROSPECT, September 2019-June 2021). As the primary outcomes of this paper are patient-caregiver discordance in prognostic information preferences and perceptions, we only included patients from the original sample with a participating caregiver.

Sample and procedure

Medical and pulmonary oncologists from seven Dutch (non)academic hospitals selected eligible patients from their outpatient clinic. Eligible patients (≥18 years old, proficient in Dutch) had incurable metastatic/inoperable cancer with an estimated median overall survival of ≤12 months at group-level (at diagnosis of advanced disease or after disease progression; with or without anticancer treatment). Physicians received an overview of all inclusion criteria, which additionally specified when a median overall survival of ≤12 months at group-level was expected per tumor type (i.e., for which type and line of anticancer treatment) based on clinical trials (Table S1, Supplemental files). Eligible patients were orally informed about the study’s focus in general terms (i.e., views on illness, treatment and prospects) and the preferred participation of a primary informal caregiver (≥18 years, proficient in Dutch). Patients were asked to invite a caregiver who was closely involved in the disease trajectory. This could be their intimate partner or, if patients did not have an intimate partner or he/she was unable or unwilling to participate, someone with whom they had a close relationship (regardless of relationship type). Interested patients received a patient information letter by e-mail or postal mail and, if applicable, a tailored version for their caregiver. Both were blinded to prognostic eligibility criteria. After providing written informed consent, patients and caregivers each completed one survey, which was sent by e-mail or postal mail. The research team instructed patients and caregivers to complete the surveys separately from each other. Study invitation was not linked to a specific moment in the disease trajectory; timing of participation could therefore vary. Procedures adhered to the Helsinki Declaration. All medical ethics review boards granted exemption from formal approval (W19_051#19.073). The predetermined sample size met general rules of thumb (≥10 observations per independent variable) and was sufficiently large to establish proportions of discordance. ⁽³⁰⁾

Measures

Surveys were composed by the research team (i.e., medical oncologists and communication researchers) and pilot tested among n=8 patients and n=8 caregivers from the target population to optimize comprehensibility, emotional load and length.

Descriptive variables

Patients and caregivers reported their sex (male/female), age, education (low/medium/high), nationality (Dutch/other) and religiosity (yes/no). Physicians reported patients’ date of diagnosis of metastatic/inoperable cancer to calculate time since diagnosis, tumor type and line of systemic treatment during study participation (none/first/second/≥third). The category “none” included patients who previously received systemic treatment, and/or received non-systemic treatment (e.g., radiotherapy) during study participation, and/or would receive (non-)systemic treatment in the future.

Dependent variables

We assessed patients’ and caregivers’ prognostic information preferences regarding the likelihood of cure and mortality risk (5/2/1 year) with four items on a binary scale (yes/no). We introduced these items with the phrase: “Are you a person who wants to know” followed by, for example, “the likelihood of (the patient) dying from your (his/her) cancer within 1 year from now?”. Items were self-developed, in absence of standardized instruments, yet inspired by prior Australian studies. ^{(31, 32)} Patient-caregiver discordance in prognostic information preferences was operationalized by comparing patients’ and caregivers’ preferences (Table 1).

Table 1.

Operationalization of dyadic variables (patient-caregiver discordance in prognostic information preferences, prognostic perceptions, main source of prognostic perceptions, avoidance coping and active coping).

			Patient	Caregiver
Prognostic information preferences Likelihood of cure One item: “Are you a person who wants to know the likelihood of (the patient) being cured from your (his/her) cancer?” Mortality risk (5/2/1 year) Three items: “Are you a person who wants to know the likelihood of (the patient) dying from your (his/her) cancer within 5/2/1 year(s) from now?”	Discordance	Patient prefers prognostic information, caregiver does not	Likelihood of cure and 5-, 2-, 1-year mortality risk Answer: “Yes, I prefer to know”.	Likelihood of cure and 5-, 2-, 1-year mortality risk Answer: “No, I prefer not to know”.
		Caregiver prefers prognostic information, patient does not	Likelihood of cure and 5-, 2-, 1-year mortality risk Answer: “No, I prefer not to know”.	Likelihood of cure and 5-, 2-, 1-year mortality risk Answer: “Yes, I prefer to know”.
	Concordance	Patient and caregiver prefer prognostic information	Likelihood of cure and 5-, 2-, 1-year mortality risk Answer: “Yes, I prefer to know”.	Likelihood of cure and 5-, 2-, 1-year mortality risk Answer: “Yes, I prefer to know”.
		Patient and caregiver do not prefer prognostic information	Likelihood of cure and 5-, 2-, 1-year mortality risk Answer: “No, I prefer not to know”.	Likelihood of cure and 5-, 2-, 1-year mortality risk Answer: “No, I prefer not to know”.
Prognostic perceptions Likelihood of cure One item: “Based on your understanding about your (the patient’s) illness, your (his/her) health in general and the treatments you are (he/she is) receiving, how likely is it that your (the patient’s) cancer will be cured?” Mortality risk (5/2/1 year) Three items: “Based on your understanding about your (the patient’s) illness, your (his/her) health in general and the treatments you are (he/she is) receiving, how likely is it that you (the patient) will die from your (his/her) cancer within 5/2/1 year(s) from now?”	Discordance	Patient is optimistic, caregiver is pessimistic	Likelihood of cure Answer: “Possible (60-40%)”, “Likely (75-60%)”, “Very likely (90-75%)” or “Extremely likely (100-90%)”. 5-, 2-, 1-year mortality risk Answer: “Unlikely (25-40%)”, “Very unlikely (10-25%)” or “Extremely unlikely (0-10%)”.	Likelihood of cure Answer: “Unlikely (40-25%)”, “Very unlikely (25-10%)” or “Extremely unlikely (10-0%)”. 5-, 2-, 1-year mortality risk Answer: “Possible (40-60%)”, “Likely (60-75%)”, “Very likely (75-90%)” or “Extremely likely (90-100%)”.
		Caregiver is optimistic, patient is pessimistic	Likelihood of cure Answer: “Unlikely (40-25%)”, “Very unlikely (25-10%)” or “Extremely unlikely (10-0%)”. 5-, 2-, 1-year mortality risk Answer: “Possible (40-60%)”, “Likely (60-75%)”, “Very likely (75-90%)” or “Extremely likely (90-100%)”.	Likelihood of cure Answer: “Possible (60-40%)”, “Likely (75-60%)”, “Very likely (90-75%)” or “Extremely likely (100-90%)”. 5-, 2-, 1-year mortality risk Answer: “Unlikely (25-40%)”, “Very unlikely (10-25%)” or “Extremely unlikely (0-10%)”.
	Concordance	Patient and caregiver are pessimistic	Likelihood of cure Answer: “Unlikely (40-25%)”, “Very unlikely (25-10%)” or “Extremely unlikely (10-0%)”. 5-, 2-, 1-year mortality risk Answer: “Possible (40-60%)”, “Likely (60-75%)”, “Very likely (75-90%)” or “Extremely likely (90-100%)”.	Likelihood of cure Answer: “Unlikely (40-25%)”, “Very unlikely (25-10%)” or “Extremely unlikely (10-0%)”. 5-, 2-, 1-year mortality risk Answer: “Possible (40-60%)”, “Likely (60-75%)”, “Very likely (75-90%)” or “Extremely likely (90-100%)”.
		Patient and caregiver are optimistic	Likelihood of cure Answer: “Possible (60-40%)”, “Likely (75-60%)”, “Very likely (90-75%)” or “Extremely likely (100-90%)”. 5-, 2-, 1-year mortality risk Answer: “Unlikely (25-40%)”, “Very unlikely (10-25%)” or “Extremely unlikely (0-10%)”.	Likelihood of cure Answer: “Possible (60-40%)”, “Likely (75-60%)”, “Very likely (90-75%)” or “Extremely likely (100-90%)”. 5-, 2-, 1-year mortality risk Answer: “Unlikely (25-40%)”, “Very unlikely (10-25%)” or “Extremely unlikely (0-10%)”.
Main source of prognostic perceptions One item: “On what information source do you base your answers to the questions about your (the patient’s) prognosis mainly?”	Discordance		Patients and caregivers choose a different answer: “healthcare provider”, “family or friends”, “support group”, “internet”, “books” or “feeling or personal beliefs”.
	Concordance		Patients and caregivers choose the same answer: “healthcare provider”, “family or friends”, “support group”, “internet”, “books” or “feeling or personal beliefs”.
Avoidance coping Example item: “Avoiding difficult situations.”	Discordance a		(Very) low scores (males, ≤11; females, ≤11).	Average scores (males, 12–17; females, 12–16)
			(Very) low scores (males, ≤11; females, ≤11).	(Very) high scores (males, ≥18; females, ≥17).
			Average scores (males, 12–17; females, 12–16).	(Very) low scores (males, ≤11; females, ≤11).
			Average scores (males, 12–17; females, 12–16).	(Very) high scores (males, ≥18; females, ≥17).
			(Very) high scores (males, ≥18; females, ≥17).	(Very) low scores (males, ≤11; females, ≤11).
			(Very) high scores (males, ≥18; females, ≥17).	Average scores (males, 12–17; females, 12–16).
	Concordance a		(Very) low scores (males, ≤11; females, ≤11).	(Very) low scores (males, ≤11; females, ≤11).
			Average scores (males, 12–17; females, 12–16).	Average scores (males, 12–17; females, 12–16).
			(Very) high scores (males, ≥18; females, ≥17).	(Very) high scores (males, ≥18; females, ≥17).
Active coping Example item: “Using a direct approach to solve the problem.”	Discordance a		(Very) low scores (males, ≤14; females, ≤15).	Average scores (males, 15–20; females, 16–20).
			(Very) low scores (males, ≤14; females, ≤15).	(Very) high scores (males, ≥21; females, ≥21).
			Average scores (males, 15–20; females, 16–20).	(Very) low scores (males, ≤14; females, ≤15).
			Average scores (males, 15–20; females, 16–20).	(Very) high scores (males, ≥21; females, ≥21).
			(Very) high scores (males, ≥21; females, ≥21).	(Very) low scores (males, ≤14; females, ≤15).
			(Very) high scores (males, ≥21; females, ≥21).	Average scores (males, 15–20; females, 16–20).
	Concordance a		(Very) low scores (males, ≤14; females, ≤15).	(Very) low scores (males, ≤14; females, ≤15).
			Average scores (males, 15–20; females, 16–20).	Average scores (males, 15–20; females, 16–20).
			(Very) high scores (males, ≥21; females, ≥21).	(Very) high scores (males, ≥21; females, ≥21).

^aBased on normative scores of the general population (18-65 years) on the avoidance- and active tackling-subscales of the Utrecht Coping List (UCL), specified by gender. ⁽³⁸⁾

We measured patients’ and caregivers’ prognostic perceptions of the likelihood of cure and mortality risk (5/2/1 year) with four items. We introduced these items with the phrase: “Based on your understanding about your (the patient’s) illness, your (his/her) health in general and the treatments you are (he/she is) receiving, how likely is it that” followed by, for example, “you (the patient) will die from your (his/her) cancer within 1 year from now?”. Scores ranged from 1-7 (“extremely unlikely, 0-10%/very unlikely, 10-25%/unlikely, 25-40%/possible, 40-60%/likely, 60-75%/very likely, 75-90%/extremely likely, 90-100%”). Items were self-developed, in absence of standardized instruments, yet inspired by Cancer Care Outcomes Research and Surveillance Consortium (CanCORS) studies. ^(33–37) Patient-caregiver discordance in prognostic perceptions was operationalized by comparing patients’ and caregivers’ perceptions (Table 1).

Independent variables

In response to literature gaps, we used a dyadic perspective to predict patient-caregiver discordance in prognostic perceptions. Hence, as independent variables, we selected factors that provided information about how people experience and deal with cancer and the associated prognosis within dyads.

We assessed patients’ and caregivers’ main source of prognostic perceptions (healthcare provider/family or friends/support group/internet/books/feeling or personal beliefs), and compared answers within dyads to operationalize patient-caregiver discordance in main source of prognostic perceptions (Table 1).

We measured patients’ and caregivers’ avoidance coping and active coping with the 8-item avoidance-subscale (e.g., “Avoiding difficult situations”) and the 7-item active tackling-subscale (e.g., “Using a direct approach to solve the problem”) of the Utrecht Coping List (UCL). ⁽³⁸⁾ Items were scored (1-4, “never” to “very often”) and summed per subscale. Sum scores were classified as (very) low, average or (very) high based on normative scores among the general population (18-65 years), specified by gender. ⁽³⁸⁾ We operationalized patient-caregiver discordance in avoidance coping and active coping by comparing patients’ and caregivers’ classifications (Table 1).

Patients reported their health-related quality of life (HR-QL) with the 2-item Global Health Status-subscale of the EORTC-Quality-of-Life-Questionnaire (EORTC-QLQ-C30). Items (i.e., “How would you rate your overall (1) health and (2) quality of life during the past week?”) were scored from 1-7 (“very poor” to “excellent”). We assessed patients’ physical functioning with the 5-item physical functioning-subscale of the EORTC-QLQ-C30. Items (i.e., “Do you need help with eating, dressing, washing yourself or using the toilet?”) were scored from 1-4 (“not at all” to “very much”). All EORCT-QLQ-C30 scores were transformed to a 0-100 scale. ⁽³⁹⁾

Caregivers reported their relationship with the patient (being a partner/parent/child/sibling/uncle/aunt/cousin/friend/neighbor/other), yet we clustered answers (partner/child/other) to limit the number of statistical comparisons. We measured caregivers’ style of providing support with three subscales of the 19-item Active Engagement, Protective Buffering and Overprotection-Questionnaire (ABO): active engagement (5 items, e.g., “I ask the patient how he or she feels“), protective buffering (8 items, e.g., “I try to hide my worries about the patient”) and overprotection (6 items, e.g., “When it comes down to it, I think that the patient does not know what’s right for him or her“). ⁽⁴⁰⁾ Items were scored (1-5, “I do not think that” to “I very strongly think that”), summed and averaged per subscale.

Statistical analysis

We calculated frequencies to present the extent of patient-caregiver discordance (four categories; Table 1) in information preferences and perceptions regarding the likelihood of cure and 5-, 2- and 1-year mortality risk (aim 1 and 2).

We calculated descriptive statistics for factors that were hypothesized to relate to patient-caregiver discordance in prognostic perceptions, distinguished by concordant and discordant dyads. To explore which factors were significantly associated with patient-caregiver discordance in prognostic perceptions (aim 3), we first tested the need for multilevel binomial logistic regression analyses (intraclass correlations ≥10%). ⁽⁴¹⁾ Multilevel analyses deal with clustered observations, missing data (using maximum likelihood estimation) and multiple testing problems. ⁽⁴²⁾ Next, we built an unconditional model (i.e., without predictors) with patient-caregiver discordance in prognostic perceptions as the dependent variable (0=concordance, 1=discordance). Level 1 (dyads) and level 2 (physicians) were kept if likelihood-ratio Chi²-tests were significant (α=.05) and intraclass correlations were ≥10%. ⁽⁴¹⁾ Subsequently, we built conditional models (i.e., with predictors) by adding fixed factors one by one, which were selected on theoretical grounds (patient-caregiver discordance in prognostic information preferences, main source of prognostic perceptions, avoidance coping and active coping; patients’ HR-QL and physical functioning; caregivers’ relationship with the patient, active engagement, protective buffering and overprotection; intercorrelations r<.80). During model building, we tested variables at a liberal α-level of 20%, preventing elimination because of confounding or modification effects. We tested the resultant model at an α-level of 5%, eliminating non-significant variables one by one to simplify the final model. Analyses were performed with IBM SPSS Statistics 26.

Results

PROSPECT included 540 patients and/or caregivers (response rate 62%; Supplemental files, Figure S1), of whom 412 dyads reported prognostic information preferences and prognostic perceptions. About half of patients was male; the mean age was 64 years (range 26-90 years). Caregivers were more often female (61%; n=250/412) and somewhat younger (mean age 58, range 21-82 years). About three-quarters (77%; n=315/412) of dyads included partner-relationships (Table 2).

Table 2.

Descriptive factors of patients and caregivers.

Descriptive factors	Cronbach’s alpha a	Patients	Caregivers
		n = 412
Sex (male), % (n)		53.6 (221)	39.3 (162)
Age (years), mean ± SD		63.7 ± 11.1	58.3 ± 13.0 b
Education, % (n) c
Low		38.2 (157) d	34.7 (143)
Medium		26.8 (110) d	26.9 (111)
High		35.0 (144) d	38.3 (158)
Nationality (Dutch), % (n)		95.9 (395)	96.6 (398)
Religiosity (yes), % (n) e		40.0 (165)	36.0 (148) f
Relationship with the patient, % (n)
Caregiver is patient’s partner		76.5 (315)	76.5 (315)
Caregiver is patient’s child		13.6 (56)	13.6 (56)
Other g		10.0 (41)	10.0 (41)
Time since diagnosis (months), mean ± SD		16.9 ± 17.4 d
Line of systemic treatment during study participation
None		24.0 (98) h
First line		42.9 (175) h
Second line		20.8 (85) h
≥Third line		12.3 (50) h
Tumor type, % (n)
Lung		25.5 (105)
Pleura		6.3 (26)
Oesophagogastric		14.6 (60)
Pancreatic		7.0 (29)
Other gastrointestinal		12.1 (50)
Colorectal		2.7 (11)
Brain		12.6 (52)
Gynaecological		10.0 (41)
Soft tissue		2.7 (11)
Other (each type n<10) i		6.6 (27)
Health-related quality of life (EORTC-QLQ-C30 subscale, 0-100), mean ± SD	.88 d	63.1 ± 20.9 d
Physical functioning (EORTC-QLQ-C30 subscale, 0-100), mean ± SD	.85 j	71.3 ± 22.8 j
Avoidance coping (UCL subscale, 8-32), mean ± SD	.73 j – .72 k	15.7 ± 3.4 d	15.0 ± 3.2 k
Active coping (UCL subscale, 7-28), mean ± SD	.82 l – .83 m	19.8 ± 3.6 d	20.6 ± 3.5 k
Patient-caregiver discordance in avoidance coping (discordant), % (n)		54.7 (223) n
Patient-caregiver discordance in active coping (discordant), % (n)		52.9 (216) n
Active engagement (ABO subscale, 1-5), mean ± SD	.77 o		4.3 ± 0.5 p
Protective buffering (ABO subscale, 1-5), mean ± SD	.70 q		2.5 ± 0.5 m
Overprotection (ABO subscale, 1-5), mean ± SD	.77 b		2.4 ± 0.7 o
Preference to know the likelihood of cure (yes), % (n)		94.1 (385) j	98.3 (405)
Preference to know the 5-year mortality risk (yes), % (n)		72.0 (296) d	84.9 (349) f
Preference to know the 2-year mortality risk (yes), % (n)		71.3 (293) d	85.6 (351) r
Preference to know the 1-year mortality risk (yes), % (n)		70.1 (288) d	85.4 (351) f
Perceived likelihood of cure, % (n)
Extremely likely		2.2 (9) s	2.9 (12) r
Very likely		2.0 (8) s	3.2 (13) r
Likely		3.0 (12) s	2.9 (12) r
Possibly		11.8 (48) s	7.3 (30) r
Unlikely		15.0 (61) s	12.9 (53) r
Very unlikely		13.1 (53) s	16.6 (68) r
Extremely unlikely		53.0 (215) s	54.1 (222) r
Perceived likelihood of dying within 5 years, % (n)
Extremely unlikely		5.4 (22) t	2.2 (9) u
Very unlikely		2.5 (10) t	1.5 (6) u
Unlikely		4.4 (18) t	3.4 (14) u
Possibly		27.3 (111)t	27.8 (113) u
Likely		11.3 (46) t	10.3 (42) u
Very likely		15.2 (62) t	15.5 (63) u
Extremely likely		33.9 (138) t	39.2 (159) u
Perceived likelihood of dying within 2 years, % (n)
Extremely unlikely		7.4 (30) t	3.7 (15) u
Very unlikely		6.1 (25) t	3.9 (16) u
Unlikely		9.3 (38) t	7.6 (31) u
Possibly		33.7 (137) t	36.5 (148) u
Likely		9.6 (39) t	9.6 (39) u
Very likely		14.7 (60) t	17.2 (70) u
Extremely likely		19.2 (78) t	21.4 (87) u
Perceived likelihood of dying within 1 year, % (n)
Extremely unlikely		14.0 (57) t	8.4 (34) o
Very unlikely		11.1 (45) t	7.4 (30) o
Unlikely		10.6 (43) t	12.3 (50) o
Possibly		36.6 (149) t	41.2 (167) o
Likely		7.1 (29) t	9.1 (37) o
Very likely		8.8 (36) t	9.9 (40) o
Extremely likely		11.8 (48) t	11.6 (47) o
Main source of prognostic perceptions, % (n)
Healthcare provider v		63.9 (262) l	66.6 (273) r
Family or friends		0.2 (1) l	1.2 (5) r
Support group		1.7 (7) l	0.5 (2) r
Internet		5.4 (22) l	8.5 (35) r
Books		0.2 (1) l	0.2 (1) r
Personal beliefs		28.5 (117) l	22.9 (94)r
Patient-caregiver discordance in main source of prognostic perceptions (discordant), % (n)		33.1 (135) n

^aInterpretation: <0.50 unacceptable, 0.50-0.60 poor, 0.60-0.70 questionable, 0.70-0.80 acceptable, 0.80-0.90 good, 0.90-1.00 excellent.

^bCaregivers: n = 401/412 (11 missing).

^cLow: elementary to low vocational education. Medium: up till medium level vocational education. High: high vocational or academic education.

^dPatients: n = 411/412 (1 missing).

^eIncluding Christianity, Islam, Buddhism, Hinduism, Judaism, Humanism, spirituality and “own belief”.

^fCaregivers: n = 411/412 (1 missing).

^gIncluding caregivers who are patient’s parent, sibling, aunt, uncle, cousin, friend or neighbor or who have another type of relationship.

^hPatients: n = 408/412 (4 missing).

ⁱIncluding melanoma, head and neck, thyroid, breast, vagina, prostate, bladder, kidney, adrenal cortex, bone, carcinoid and unknown primary tumors.

^jPatients: n = 409/412 (3 missing).

^kCaregivers: n = 409/412 (3 missing).

^lPatients: n = 410/412 (2 missing).

^mCaregivers: n = 407/412 (5 missing).

ⁿPatient-caregiver dyads: n = 408/412 (4 missing).

^oCaregivers: n = 405/412 (7 missing).

^pCaregivers: n = 408/412 (4 missing).

^qCaregivers: n = 404/412 (8 missing).

^rCaregivers: n = 410/412 (2 missing).

^sPatients: n = 406/412 (6 missing).

^tPatients: n = 407/412 (5 missing).

^uCaregivers: n = 406/412 (6 missing).

^vIncluding treating physician, second opinion physician, nurse, general practitioner or other healthcare provider.

Abbreviations: n: sample size; SD: standard deviation; EORTC-QLQ-C30: European Organization for Research and Treatment of Cancer Quality of Life Questionnaire for Cancer; UCL: Utrecht Coping List; ABO: Active Engagement, Protective Buffering and Overprotection.

Patient-caregiver discordance in prognostic information preferences (aim 1)

Few patient-caregiver dyads had discordant information preferences regarding the likelihood of cure (7%; n=29/409); both dyad-members generally wanted to know. In contrast, 24-25% of dyads had discordant information preferences regarding the 5-year (n=104/410), 2-year (n=99/409) or 1-year (n=102/410) mortality risk (Table 3). Overall, dyads with discordant prognostic information preferences mostly encompassed caregivers with and patients without a wish to know (75-80%; n_{likelihood of cure}=23/29, n_{5-year mortality}=78/104, n_{2-year mortality}=78/99, n_{1-year mortality}=82/102). Dyads with concordant preferences mostly encompassed patients and caregivers who both wanted prognostic information (87-100%; n_{likelihood of cure}=379/380, n_{5-year mortality}=270/306, n_{2-year mortality}=272/310, n_{1-year mortality}=268/308).

Table 3.

Discordance and concordance between patients’ and caregivers’ information preferences regarding prognosis.

Prognostic information preferences	Likelihood of cure % (n)	5-Year mortality risk % (n)	2-Year mortality risk % (n)	1-Year mortality risk % (n)
Discordance	7.1 (29/409)	25.4 (104/410)	24.2 (99/409)	24.9 (102/410)
Patient prefers prognostic information, caregiver does not	1.5 (6/409)	6.3 (26/410)	5.1 (21/409)	4.9 (20/410)
Proportion within discordant dyads	20.7 (6/29)	25.0 (26/104)	21.2 (21/99)	19.6 (20/102)
Caregiver prefers prognostic information, patient does not	5.6 (23/409)	19.0 (78/410)	19.1 (78/409)	20.0 (82/410)
Proportion within discordant dyads	79.3 (23/29)	75.0 (78/104)	78.8 (78/99)	80.4 (82/102)
Concordance	92.9 (380/409)	74.6 (306/410)	75.8 (310/409)	75.1 (308/410)
Patient and caregiver prefer prognostic information	92.7 (379/409)	65.9 (270/410)	66.5 (272/409)	65.4 (268/410)
Proportion within concordant dyads	99.7 (379/380)	88.2 (270/306)	87.7 (272/310)	87.0 (268/308)
Patient and caregiver do not prefer prognostic information	0.2 (1/409)	8.8 (36/410)	9.3 (38/409)	9.8 (40/410)
Proportion within concordant dyads	0.3 (1/380)	11.8 (36/306)	12.3 (38/310)	13.0 (40/308)
Total sample	100% (409)	100% (410)	100% (409)	100% (410)

Abbreviations: n: sample size.

Patient-caregiver discordance in prognostic perceptions (aim 2)

About 17% of patient-caregiver dyads (n=67/404) had discordant perceptions of the likelihood of cure. Discordance in perceptions of mortality risk varied between 12-25% (n_{5-year mortality}=47/401, n_{2-year mortality}=76/402, n_{1-year mortality}=101/401). The shorter the indicated survival time (5/2/1 year), the larger the proportion of dyads with discordant perceptions (Table 4). Overall, dyads with discordant perceptions mostly encompassed patients who were more optimistic about prognosis than caregivers (57-70%; n_{likelihood of cure}=38/67, n_{5-year mortality}=33/47, n_{2-year mortality}=53/76, n_{1-year mortality}=65/101). Dyads with concordant perceptions mostly encompassed patients and caregivers who both were pessimistic about patients’ prognosis (75-96%; n_{likelihood of cure}=299/337, n_{5-year mortality}=340/354, n_{2-year mortality}=288/326, n_{1-year mortality}=225/300). Still, our findings on concordance suggest that the proportion of dyads with optimistic perceptions is larger when the indicated survival time is shorter (5/2/1 year).

Table 4.

Discordance and concordance between patients’ and caregivers’ perceptions of prognosis.

Prognostic perceptions	Likelihood of cure % (n)	5-Year mortality risk % (n)	2-Year mortality risk % (n)	1-Year mortality risk % (n)
Discordance	16.6 (67/404)	11.7 (47/401)	18.9 (76/402)	25.2 (101/401)
Patient is optimistic, caregiver is pessimistic	9.4 (38/404)	8.2 (33/401)	13.2 (53/402)	16.2 (65/401)
Proportion within discordant dyads	56.7 (38/67)	70.2 (33/47)	69.7 (53/76)	64.4 (65/101)
Caregiver is optimistic, patient is pessimistic	7.2 (29/404)	3.5 (14/401)	5.7 (23/402)	9.0 (36/401)
Proportion within discordant dyads	43.3 (29/67)	29.8 (14/47)	30.3 (23/76)	35.6 (36/101)
Concordance	83.4 (337/404)	88.3 (354/401)	81.1 (326/402)	74.8 (300/401)
Patient and caregiver are pessimistic	74.0 (299/404)	84.8 (340/401)	71.6 (288/402)	56.1 (225/401)
Proportion within concordant dyads	88.7 (299/337)	96.0 (340/354)	88.3 (288/326)	75.0 (225/300)
Patient and caregiver are optimistic	9.4 (38/404)	3.5 (14/401)	9.5 (38/402)	18.7 (75/401)
Proportion within concordant dyads	11.3 (38/337)	4.0 (14/354)	11.7 (38/326)	25.0 (75/300)
Total	100% (404)	100% (401)	100% (402)	100% (401)

Abbreviations: n: sample size.

Factors associated with patient-caregiver discordance in prognostic perceptions (aim 3)

Table 5 describes the factors that were hypothesized to relate to patient-caregiver discordance in prognostic perceptions, for concordant and discordant dyads separately. Multilevel analyses showed that patient-caregiver dyads with discordant prognostic information preferences were significantly more likely to have discordant perceptions of the 1-year mortality risk than dyads with concordant information preferences (OR=1.66, 95%CI [1.02; 2.69], p<.05; Table 6). Additionally, dyads in which patients had better physical functioning were significantly more likely to have discordant perceptions of the 1-year mortality risk (OR=1.01, 95%CI [1.00; 1.02], p<.01). There were no significant associations of patient-caregiver discordance in prognostic perceptions with discordance in the main source of prognostic perceptions, avoidance coping or active coping, nor with patients’ health-related quality of life or caregivers’ relationship with the patient (partner/child/other), active engagement, protective buffering or overprotection.

Table 5.

Descriptive statistics of factors potentially associated with patient-caregiver discordance in prognostic perceptions (included in multilevel analyses).

Factor	Patient-caregiver dyads with discordant perceptions of the 1-year mortality risk n = 99	Patient-caregiver dyads with concordant perceptions of the 1-year mortality risk n = 298
Relationship with the patient, % (n)
Caregiver is patient’s partner	77.8 (77)	75.8 (226)
Caregiver is patient’s child	12.1 (12)	14.1 (42)
Other a	10.1 (10)	10.1 (30)
Health-related quality of life (EORTC-QLQ-C30 subscale, 0-100), mean ± SD	66.3 ± 19.5	61.7 ± 21.5
Physical functioning (EORTC-QLQ-C30 subscale, 0-100), mean ± SD	76.4 ± 17.3	69.9 ± 24.2
Patient-caregiver discordance in avoidance coping (discordant), % (n)	49.0 (48) b	56.9 (169) b
Patient-caregiver discordance in active coping (discordant), % (n)	46.9 (46) b	54.5 (162) b
Active engagement (ABO subscale, 1-5), mean ± SD	4.3 ± 0.5 b	4.4 ± 0.5 c
Protective buffering (ABO subscale, 1-5), mean ± SD	2.5 ± 0.6 c	2.4 ± 0.5 c
Overprotection (ABO subscale, 1-5), mean ± SD	2.4 ± 0.7 b	2.4 ± 0.7 d
Patient-caregiver discordance in preference to know the 1-year mortality risk (discordant), % (n)	32.3 (32)	22.1 (66)
Patient-caregiver discordance in main source of prognostic perceptions (discordant), % (n)	41.4 (41)	30.6 (91) b

^aIncluding caregivers who are patient’s parent, sibling, aunt, uncle, cousin, friend or neighbor or who have another type of relationship.

^bPatient-caregiver dyads with discordant prognostic perceptions: n = 98/99 (1 missing). Patient-caregiver dyads with concordant prognostic perceptions: n = 297/298 (1 missing).

^cPatient-caregiver dyads with discordant prognostic perceptions: n = 97/99 (2 missing). Patient-caregiver dyads with concordant prognostic perceptions: n = 296/298 (2 missing).

^dPatient-caregiver dyads with concordant prognostic perceptions: n = 293/298 (5 missing).

Abbreviations: n: sample size; SD: standard deviation; EORTC-QLQ-C30: European Organization for Research and Treatment of Cancer Quality of Life Questionnaire for Cancer; ABO: Active Engagement, Protective Buffering and Overprotection.

Table 6.

Mixed-effects binomial logistic regression model with predictors of patient-caregiver discordance in perceptions of the 1-year mortality risk. ^a

	Final model b n = 397 c
Patient-caregiver discordance in perceptions of the 1-year mortality risk	B	p	Exp(b)	Lower CI Exp(b)	Upper CI Exp(b)
Intercept	−2.228	.000***	.108	.050	.233
Patients’ physical functioning (EORTC-QLQ-C30)	.013	.008**	1.013	1.003	1.022
Patient-caregiver dyads’ information preference for 1-year mortality risk (ref = concordant) d	.506	.040*	1.659	1.022	2.692

^a0 = Concordance between patients’ and caregivers’ perceptions of prognosis, 1 = Discordance between patients’ and caregivers’ perceptions of prognosis.

^bWe used a mixed-effects binomial logistic regression model with two levels, because of an ICC = 9.8%. Intercorrelations between predictors were <.54. Patient-caregiver discordance in active coping and caregivers’ relationship with the patient, active engagement and protective buffering were omitted from the model (p>.20). To simplify the final model, patients’ health-related quality of life, patient-caregiver discordance in avoidance coping, caregivers’ overprotection and patient-caregiver discordance in main source of prognostic perceptions (p>.05) were eliminated. Associations of patients’ physical functioning and patient-caregiver discordance in prognostic information preferences with patient-caregiver discordance in prognostic perceptions remained significant in models that controlled for caregivers’ relationship with the patient.

^cPatient-caregiver dyads: n = 397/412 (11 missing on dependent variable, 4 missing on independent variables).

^d0 = Concordance between patients’ and caregivers’ prognostic information preference, 1 = Discordance between patients’ and caregivers’ prognostic information preference.

^*Significant at p<.05.

^**Significant at p<.01.

^***Significant at p<.001.

Abbreviations: n: sample size; B: unstandardized coefficient; p: significance; Exp(b): exponentiation of the B coefficient, which is an odds ratio; CI Exp(b): 95% Confidence Interval of exponentiation of the B coefficient; ref: reference category; EORTC-QLQ-C30: European Organization for Research and Treatment of Cancer Quality of Life Questionnaire for Cancer; ICC: intraclass correlation.

Discussion

This is the first study among patient-caregiver dyads to reveal that up to 25% have discordant preferences for information about patients’ mortality risk. Like most Western literature on dyad-members’ individual information preferences suggests, we observed that patients are less likely to want prognostic information than caregivers. ^{(13, 43)} According to prior studies, patients’ and caregivers’ readiness to deal with prognostic information could differ. ^{(44, 45)} Being confronted with one’s own mortality is undeniably confrontational, likely triggering coping mechanisms and emotions. For some patients, prognostic ignorance is existentially essential to endure sickness and treatment. ⁽⁴⁶⁾ It is plausible that patients’ level of acceptance of the terminal diagnosis affects their desire for prognostic information. Caregivers, however, possibly approach prognostic information more instrumentally, as they are managing patient’s care and daily tasks as well as their own future. ^{(23, 47)}

Up to one-fourth of patient-caregiver dyads had discordant prognostic perceptions. Such discordance was more likely when predicting death within a shorter timeframe (5/2/1 year). Dyad-members potentially felt most uncertain about the 1-year mortality risk, leaving room for diverging perceptions. Considering the median survival of ≤12 months in our sample, death within five years is more plausible and therefore perhaps harder to disagree on. Alternatively, our measures captured psychological processes (e.g., disbelief, hope, sadness) rather than knowledge of prognosis only. These psychological processes –affecting prognostic perceptions– might be most influential when confronted with closer proximity to death. In line with literature, we showed that caregivers are generally more pessimistic about prognosis than patients. ^{(20, 21)} Caregivers’ “outsider” perspective possibly facilitates more objective prognostication. Interestingly, previous research points out that cognitive prognostic understanding does not necessarily imply emotional preparedness for death, and hope can exist alongside recognition of terminal cancer. ^{(9, 48, 49)}

Inspecting both prognostic information preferences and prognostic perceptions, it is remarkable that nearly all dyads agreed on wanting to know the likelihood of cure, yet they relatively often had different understandings of advanced cancers’ curability. Potentially, hope for any chance of cure continues to live as long as patients receive treatment. ^{(50, 51)} Contrarily, one-quarter of dyads disagreed on wanting 5-year mortality risk information, yet they relatively often agreed on the likelihood of dying within this timeframe. When contemplating a fairly distant future, patients and caregivers seemingly both adopt a more realistic perspective, independent of wanting prognostic information. These hypotheses are speculative; additional research needs to clarify how and why patients and caregivers differentially perceive various types of prognostic information.

We are the first to demonstrate that discordant prognostic information preferences among patient-caregiver dyads could impede a shared understanding of prognosis. We did not find associations of discordant information sources with discordant prognostic perceptions, suggesting that agreement about if dyad-members want prognostic information is more influential than acquiring information from the same source. However, causality of the effect of discordant information preferences cannot be inferred. It could also be that dyads who think about patients’ prospects differently therefore have different preferences for prognostic information. Associations with discordance in personal coping styles (i.e., active and avoidance coping) or caregivers’ styles of support (i.e., active engagement, protective buffering, overprotection) were absent; possibly these were accounted for by the effect of discordant prognostic information preferences. An alternative explanation might be that our measures assessed reactions to problems in general – not cancer or prognosis in particular. Although patients’ overall HR-QL had no influence, we showed that patient-caregiver dyads are more likely to perceive the 1-year mortality risk discordantly when patient’s physical functioning is relatively good. It is conceivable that prognostication is more difficult when indicators of physical deterioration are lacking, and patients still participate in daily activities. ⁽²⁹⁾

Discordance in prognostic information preferences and perceptions could complicate patient-caregiver communication about the disease and the future (and vice versa). When caregivers who prefer prognostic information cannot discuss prognosis with patients who favor ignorance, it is challenging to prepare emotionally and practically (e.g., planning a funeral) for the end-of-life together. ^{(2, 5, 6, 22)} For caregivers specifically, being unprepared for patient’s death may contribute to complicated grief during bereavement. ^{(52, 53)} Also, discordance among patient-caregiver dyads adds complexity to clinical interactions. Physicians may need to navigate conflicting prognostic information needs and balance differing prognostic perceptions. Deliberating discontinuation of aggressive anticancer treatment, for example, seems difficult if one dyad-member perceives prognosis optimistically. ^{(10, 14, 19)}

Open dyadic communication could increase intimacy, satisfaction with the relationship and patient care, and reduce interpersonal conflict. ^{(2, 5–8, 11)} For physicians, it might be helpful to explain that prognostic information preferences and perceptions can differ from person to person. Physicians may explicitly ask what patients and caregivers (want to) know about prognosis and explore if and why there are differences. Such communication allows awareness of discordance and deliberation of how this could be handled. While acknowledging both dyad-members’ viewpoints, physicians may explicate their primary responsibility for meeting patients’ information needs, given the therapeutic relationship. ⁽⁵⁴⁾ If desired, accurate prognostic estimates should be delivered empathetically; adding multiple survival scenarios (e.g., best- and worst-case) might help balancing realism with hope and allows coexistence of different prognostic perceptions. ^{(24, 55)} If patients reject prognostic discussions, caregivers could share their thoughts with other supportive people, which is shown to relieve distress in both dyad-members. ⁽⁸⁾ These complex situations nonetheless raise ethical –and culturally sensitive– questions about the degree to which physicians’ responsibility extends beyond the patient. ⁽⁵⁴⁾ Besides tailoring prognostic information, physicians may highlight that dyadic open communication has emotional and practical benefits. To better understand dyads’ communication patterns and ultimately guide them in coping with cancer together, assistance of a nurse or medical psychologist seems advisable. ⁽⁸⁾ In this regard, policy makers may focus on developing couple-focused interventions to stimulate communication about dyadic coping with cancer, including communication about differences in information needs and prognostic perceptions. Previous work demonstrated that simple tools such as question prompt lists can promote prognostic discussions among advanced cancer patients, caregivers and physicians. ⁽⁵⁶⁾

Our research has limitations. First, our measures of prognostic information preferences and prognostic perceptions were non-validated. Second, our operationalization of discordance in prognostic perceptions (i.e., dichotomization of 7-point scales) implies that there is variation regarding how far apart dyad-members’ discordant estimates are. Yet, we chose this method to reveal if patient-caregiver dyads agreed about prognosis on a basic, “gist” level, preventing overestimation of discordance rates by incorporating minor differences. Besides, we cannot draw conclusions about the accuracy of prognostic perceptions, as we did not include physicians’ estimates or actual survival data in our analyses. Third, our secondary analyses did not include measurements of patient-caregiver communication about prognosis or physicians’ prognostic (non-)disclosure to dyads, which both may contribute to patient-caregiver discordance in prognostic perceptions in reality. Last, the study’s Dutch setting restricts generalizability and its cross-sectional nature cannot determine causality.

Future quantitative research may investigate the unique drivers of patients’ or caregivers’ more optimistic prognostic perceptions, considering individual coping styles, fighting spirit, death acceptance and denial as predictors. Longitudinal survey studies could reveal if, how and why patients’ and caregivers’ shared understanding of prognosis may change over time. Qualitative research (e.g., ethnography, interviews) may uncover how dyads communicate about prognosis, whether they are aware of possibly discordant prognostic perceptions, and how dyads deal with prognostic discordance. Lastly, while this study did not find associations with relationship type, it would be interesting to explore if prognostic communication is different within dyads who are intimate partners opposed to other loved ones.

Strengths include the investigation of prognostic information preferences and perceptions within patient-caregiver dyads rather than individuals, among a relatively large sample. We gained thorough insight into patients’ and caregivers’ shared understanding of prognosis by examining multiple aspects of prognosis. Finally, we are the first to study dyadic factors that predict patient-caregiver discordance in prognostic perceptions, which provides entry points for interventions to help dyads deal with advanced cancer.

Conclusions

A substantial proportion of patient-caregiver dyads in advanced cancer care have discordant preferences for prognostic information. Patient-caregiver dyads with discordant prognostic information preferences and dyads including a patient with better physical functioning are more likely to hold discordant prognostic perceptions. Physicians should gauge patients’ and caregivers’ prognostic information preferences and prognostic perceptions, and address the potential discordance.

Key Message.

This cross-sectional survey study in advanced cancer care shows that up to one-quarter of patient-caregiver dyads have discordant prognostic information preferences, and –similarly– up to one-quarter perceive patients’ prognosis discordantly. Hence, physicians should explore patients’ and caregivers’ prognostic information preferences and prognostic perceptions, and be sensitive to potential discordance.