Prognostic disclosure, end-of-life communication, and social support associated with distinct dyadic death-preparedness states among cancer dyads

Abstract

Background

Preparing for death is relational because patients with advanced cancer and family caregivers confront approaching death together. We examined individual and relational factors associated with distinct dyadic death-preparedness states during patients’ final 6 months of life.

Patients and Methods

In this longitudinal cohort study of 694 patient-caregiver dyads, hierarchical generalized linear models with lagged time-varying predictors examined unprepared-concordant, cognitive-concordant, emotional-concordant, and sufficient-concordant states versus discordance across individual, relationship, and relationship-process domains.

Results

Older patient age was associated with lower likelihood of cognitive concordance but higher likelihood of sufficient concordance, and higher patient educational attainment was associated with sufficient concordance. Patient- and caregiver-reported physician prognostic disclosure was associated with higher likelihoods of cognitive concordance (AORs, 3.416 and 2.411, respectively) and sufficient concordance (AORs, 3.743 and 2.854, respectively). Greater patient-reported end-of-life discussions with caregivers were associated with a higher likelihood of cognitive concordance (AOR, 1.190; 95% CI, 1.106-1.279) but a lower likelihood of emotional concordance (AOR, 0.800; 95% CI, 0.650-0.985). Higher patient- and caregiver-reported social support was associated with a lower likelihood of cognitive concordance (AORs, 0.956 and 0.945, respectively).

Conclusions

Relationship-process factors were more consistently associated with dyadic death-preparedness states than individual or relationship characteristics. These findings support assessing both patients’ and caregivers’ cognitive and emotional preparedness and considering reported prognostic disclosure, patient-caregiver end-of-life discussions, and social-support dynamics when evaluating dyadic preparation for death.

Implications for practice.

Death preparedness should be assessed in both patients with advanced cancer and family caregivers because shared prognostic understanding does not necessarily indicate shared emotional readiness. Relationship-process factors—particularly participant-reported physician prognostic disclosure, patient-reported EOL discussions with caregivers, and perceived social support—were more consistently associated with dyadic preparedness than individual or relationship characteristics. Clinicians should assess cognitive and emotional preparedness separately in both members, consider their reports and understanding of prognostic communication, elicit both perspectives on EOL discussions, and explore how family and broader social networks participate in prognostic communication.

Introduction

Dying from cancer affects both patients and their family caregivers (hereafter, caregivers).¹ In health care, a dyadic approach considers the patient and a closely involved person, such as a family caregiver, as an interdependent unit whose experiences and responses may reciprocally influence one another.¹ Both may experience death anxiety and anticipatory grief as they confront the patient’s approaching death.^2,3 Preparing for death—a process involving cognitive prognostic awareness and emotional acceptance—helps patients and caregivers adapt to the dying trajectory and engage in end-of-life (EOL) planning.^4,5 For clinicians caring for patients with advanced cancer, supporting such preparation requires attention not only to individual patients but also to how patients and caregivers understand, communicate about, and emotionally respond to approaching death. Death preparedness is therefore both an individual and relational process because patients’ and caregivers’ experiences unfold within an interdependent relationship.^6,7 However, they may differ in how they process prognostic information, communicate about impending death, cope with illness, and adjust emotionally,¹ giving rise to different patterns of shared death preparedness.

Patient-caregiver concordance reflects whether both members share the same death-preparedness state. However, concordance is often limited and increases only during the final month of life.⁸ Moreover, concordance is not inherently adaptive; its implications depend on the preparedness state in which agreement occurs. Dyadic death preparedness comprises 5 mutually exclusive states: the unprepared-concordant, cognitive-concordant, emotional-concordant, sufficient-concordant, and discordant states.⁸ The unprepared-concordant state indicates that both patients and caregivers lack accurate prognostic awareness and emotional preparedness, whereas the sufficient-concordant state reflects shared cognitive understanding and emotional readiness. The cognitive- and emotional-concordant states indicate agreement in only 1 component, whereas the discordant state reflects disagreement in cognitive and/or emotional preparedness.

Different dyadic death-preparedness states have distinct psychological and clinical implications. Compared with the discordant state, the sufficient-concordant state is associated with fewer patient anxiety symptoms, fewer caregiver depressive symptoms, better dyadic quality of life (QOL),⁹ lower likelihoods of chemotherapy and aggressive life-sustaining treatments, and greater hospice use.¹⁰ In contrast, patients in the cognitive-concordant state report greater anxiety and poorer QOL, whereas those in the emotional-concordant state report fewer anxiety and depressive symptoms.⁹ Caregivers in the unprepared- and cognitive-concordant states also report greater depressive symptoms and poorer QOL.⁹ Thus, neither concordance nor discordance is uniformly adaptive; their implications depend on the preparedness state. Identifying factors associated with these states may therefore improve understanding of the relational context in which patients and caregivers prepare for approaching death, engage in EOL planning,¹¹ and pursue a good death.¹²

Because patients’ and caregivers’ understanding, emotional responses, and communication are interdependent throughout serious illness, relationship processes, including prognostic disclosure, EOL communication, and social support, may be particularly relevant to shared death preparedness. Systematic reviews indicate that dyadic interventions in cancer and advance care planning can improve patient or caregiver outcomes¹³ and facilitate EOL communication and decision-making.¹⁴ However, benefits are more consistently demonstrated for individual than dyadic outcomes,¹⁵ and few interventions explicitly address patient-caregiver alignment.¹⁶ For oncology practice, identifying relational processes associated with different forms of shared preparedness may improve understanding of the relational context in which patients and caregivers approach prognostic and EOL communication. Although our previous studies described trajectories of dyadic concordance,⁸ its psychological and clinical consequences,^9,10 and predictors of preparedness among patients¹⁷ and caregivers¹⁸ separately, they did not examine factors associated with dyadic death preparedness itself. Because dyadic concordance is a relational construct that cannot be inferred from individual-level analyses, identifying factors associated with dyadic concordance addresses a distinct scientific and clinical question.

Existing reviews have focused on related outcomes, including dyadic concordance in QOL assessments across the cancer trajectory¹⁹ and EOL decision-making among patients receiving palliative care.¹⁶ Others have emphasized family communication and social support in relation to prognostic awareness.^20,21 Three methodological limitations remain. First, dyadic concordance is typically treated as static because most studies are cross-sectional.^16,19–22 Second, concordance is commonly operationalized as binary (concordant vs discordant), overlooking distinct dyadic death-preparedness states. Third, potential correlates have been examined inconsistently and rarely within an explicit dyadic framework. The Dyadic Cancer Outcomes Framework (DCOF) proposes that dyadic outcomes arise from interactions among individual characteristics, relationship characteristics, relationship processes, and broader social context.¹ Guided by the DCOF and prior theory and empirical evidence,^16–22 we longitudinally examined associations of individual characteristics, relationship characteristics, and relationship processes, including participant-reported physician prognostic disclosure, patient-caregiver EOL communication, and perceived social support, with distinct dyadic death-preparedness states during patients’ final 6 months of life.

Methods

Research design and sample

This prospective longitudinal cohort study examined factors associated with distinct dyadic death-preparedness states among cancer patient-caregiver dyads. Detailed descriptions of the parent cohort and study measures have been reported previously.^8,9

Two successive cohorts of adult patients with cancer and their primary family caregivers were recruited by convenience sampling from a tertiary medical center in northwest Taiwan between August 2015 and December 2022, with follow-up through June 2024. Eligible patients had cancer deemed by their oncologist to be progressive and unresponsive to repeated chemotherapy or immunotherapy. Caregivers were the primary providers of unpaid care and were able to participate in study interviews; no additional exclusion criteria were applied. Patients and caregivers were interviewed separately at approximately monthly intervals until patient death or study withdrawal, yielding repeated observations for each dyad. Of 746 enrolled dyads, 694 contributed dyadic death-preparedness outcome data and comprised the study sample. The longitudinal analyses were restricted to assessments occurring during patients’ final 6 months of life. The Chang Gung Medical Foundation Institutional Review Board approved both studies (103-7015B and 104-9388B), which were conducted in accordance with the Declaration of Helsinki. Written informed consent was obtained from all participants. This study followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guideline.

Measures

Outcome variable: dyadic death-preparedness states

Cognitive prognostic awareness

At each assessment, participants were asked whether they knew the prognosis and, if so, whether the cancer (1) was curable, (2) might recur but was not currently life-threatening, or (3) was not curable and death was likely imminent. Response 3 indicated accurate prognostic awareness; all other responses indicated inaccurate awareness. This measure assessed participants’ own understanding of prognosis rather than whether prognostic information had been disclosed by a physician.

Emotional preparedness for death

Emotional preparedness was assessed using 5 items from the Preparation for End-of-Life subscale of the Quality of Life at the End of Life scale.^23,24 Items were rated on a 5-point Likert scale and addressed concerns such as financial burden on the family, preparation for death, fear of dying, and life regrets (for patients). Scores ranged from 5 to 25, with higher scores indicating greater preparedness. Emotional preparedness was dichotomized using the sample median (≥20 vs <20), consistent with prior work,²⁵ and used with prognostic awareness as an indicator in latent transition analysis.

Death-preparedness states

Based on prognostic awareness and emotional preparedness, 4 death-preparedness states were previously identified for patients and caregivers separately: no preparedness, cognitive preparedness, emotional preparedness, and sufficient preparedness.⁸ The no-preparedness state indicates neither accurate prognostic awareness nor emotional readiness, whereas the sufficient-preparedness state includes both. The cognitive- and emotional-preparedness states reflect the presence of only 1 component. Supplementary Methods summarize state identification and transitions over time. At each assessment, patients and caregivers were assigned to the state with the highest posterior probability from the previously estimated latent transition model.

Dyadic death-preparedness states

Dyadic death-preparedness states were derived by cross-tabulating patient and caregiver death-preparedness states and classified as the unprepared-, cognitive-, emotional-, or sufficient-concordant states; all remaining combinations comprised the discordant state (Table S1).

Independent variables

Within the DCOF,¹ outcomes are shaped by individual (patient and caregiver) characteristics, relationship (dyad-level) characteristics, relationship processes, and social context. The social-context domain was not included because the parent cohorts did not comprehensively assess broader cultural, religious, or health-system contextual factors.

Previous analyses of this cohort showed that dyadic death preparedness was associated with patient (eg, anxiety, depression, and QOL) and caregiver (eg, caregiving burden, depressive symptoms, and QOL) outcomes.⁹ Because these variables were conceptualized as outcomes of death preparedness in those analyses rather than antecedent factors in the DCOF-guided model tested here, they were excluded to avoid overadjustment and preserve temporal ordering.

Patient characteristics

Sociodemographic variables included gender, age, marital status (ever married/not single vs unmarried/single), and educational attainment (> junior high school/≤ junior high school). Disease-related variables included the presence of any of 20 chronic conditions (yes/no), time since diagnosis (months), symptom distress, and functional dependence. Symptom distress was assessed using the 13-item Symptom Distress Scale (range, 13-65; higher scores indicate greater distress).²⁶ Functional dependence was measured using the 10-item Enforced Social Dependency Scale (range, 10-51; higher scores indicate greater dependence).²⁷

Caregiver characteristics

Sociodemographic variables included gender, age, marital status (ever married/not single vs unmarried/single), and educational attainment (≥ high school/< high school).

Relationship characteristics

The patient-caregiver relationship was categorized as spouse, adult child, or other.

Relationship-process factors

Consistent with the DCOF,¹ physician prognostic disclosure, patient-caregiver EOL discussions, and perceived social support were conceptualized as relationship-process factors because they arise through ongoing interactions among clinicians, patients, caregivers, and their social networks during serious illness.

Participant-reported physician prognostic disclosure was assessed at each interview. At baseline, participants reported whether a physician had previously disclosed that the disease was incurable (yes vs no); at each follow-up, they reported whether such disclosure had occurred since the preceding interview (yes vs no). Responses were coded according to each assessment-specific report and were not carried forward from earlier interviews.

Patient-caregiver EOL discussions were assessed using the 5-item measure developed by Steinhauser and colleagues,²⁸ which evaluates discussions regarding (1) disease progression, (2) possible EOL care, (3) EOL-care preferences, (4) unfinished business, and (5) preparation for the patient’s permanent absence. Items were rated from 1 (not at all) to 5 (always) and summed to yield a total score ranging from 5 to 25, with higher scores indicating greater patient-caregiver EOL discussion.

Perceived social support was assessed using the 19-item Medical Outcomes Study Social Support Survey (MOS-SSS),²⁹ which measures tangible, emotional, affectionate, and informational support and positive social interaction from the broader social network. Scores were transformed to a 0-100 scale, with higher scores indicating greater support.

Statistical analysis

Factors associated with dyadic death-preparedness states were examined using hierarchical generalized linear models with a multinomial logit link,³⁰ with the discordant state as the reference outcome. Random intercepts accounted for correlations among repeated observations within dyads. Both predictor and outcome assessments were restricted to patients’ final 6 months of life. For the lagged analyses, an outcome observation was included only when a preceding assessment within this 6-month period was available; measures from that preceding assessment served as predictors of the subsequent dyadic death-preparedness state.

Time-varying variables—symptom distress, functional dependence, participant-reported physician prognostic disclosure, patient-caregiver EOL discussions, and perceived social support—were lagged by 1 assessment relative to the subsequent dyadic death-preparedness state to strengthen temporal ordering. Models included the prespecified patient, caregiver, relationship, and relationship-process factors described above and were adjusted for time until death. Regression coefficients were exponentiated to obtain adjusted odds ratios (AORs) with 95% confidence intervals (CIs). Candidate predictors were selected a priori based on the DCOF^¹ and prior empirical evidence,^16–22 rather than through data-driven variable selection. Analyses were performed using IBM SPSS Statistics for Windows, version 22.0 (IBM Corp., Armonk, NY, USA).

Results

Participant characteristics and distribution of dyadic death-preparedness states

Baseline characteristics of the 694 cancer patient-caregiver dyads are shown in Table 1; most patients were male, whereas most caregivers were female and spouses. Across follow-up, patients and caregivers completed a mean (SD) of 5.30 (6.69) and 5.51 (6.49) assessments, respectively, with final assessments occurring a mean (SD) of 25.0 (24.5) and 17.0 (19.0) days before death, respectively.⁸ Consistent with our prior report,⁸ the distribution of dyadic death-preparedness states varied as death approached (Table 1 and Table S1). Accordingly, the multivariable models adjusted for time until death, with corresponding estimates presented in Table S2.

Table 1.

Baseline participant characteristics by dyadic death-preparedness state and state prevalence during the last 6 months of life.

Dyadic death-preparedness states		Unprepared-concordant			Cognitive-concordant		Emotional-concordant		Sufficient-concordant		Discordant		Total
Variable		n	%		n	%	n	%	n	%	n	%	N	%
Time until death
1 month		5		1.0	120	23.9	2	0.4	92	18.3	284	56.5	503	72.5
2 months		6		1.3	89	18.8	8	1.7	78	16.5	292	61.7	473	68.2
3 months		5		1.3	63	16.8	5	1.3	73	19.4	230	61.2	376	54.2
4 months		2		0.8	31	12.2	4	1.6	50	19.6	168	65.9	255	36.7
5 months		5		2.3	25	11.6	3	1.4	29	13.5	153	71.2	215	31.0
6 months		5		2.9	14	8.0	8	4.6	28	16.1	119	68.4	174	25.1
Individual-level factors
Patient socio-demographics and disease characteristics
Gender
Male		14		87.5	88	80.7	14	82.4	78	70.9	333	75.3	527	75.9
Female		2		12.5	21	19.3	3	17.6	32	29.1	109	24.7	167	24.1
Age, years; mean (SD)		58.8 (7.9)			55.9 (9.1)		58.9 (10.1)		63.4 (9.2)		60.2 (10.7)		60.0 (10.4)
Marital status
Ever married/not single		15		93.8	102	93.6	17	100.0	105	95.5	421	95.2	660	95.1
Unmarried/single		1		6.3	7	6.4	0	0.0	5	4.5	21	4.8	34	4.9
Educational attainment
>Junior high school		5		31.3	61	56.0	4	23.5	52	47.3	167	37.9	289	41.7
≤Junior high school		11		68.8	48	44.0	13	76.5	58	52.7	274	62.1	404	58.3
Time since diagnosis (months); mean (SD)		4.8 (5.5)			14.8 (21.6)		5.9 (8.6)		13.9 (16.3)		13.2 (22.2)		13.2 (20.8)
Symptom distress; mean (SD)		26.9 (6.0)			28.0 (6.6)		24.5 (5.2)		26.5 (6.8)		27.0 (6.4)		27.0 (6.5)
Functional dependence; mean (SD)		29.5 (6.8)			29.1 (8.0)		23.6 (6.9)		27.7 (7.9)		27.9 (8.0)		28.0 (8.0)
Other chronic disease
Yes		13		81.3	72	66.1	14	82.4	78	70.9	296	67.0	473	68.2
No		3		18.8	37	33.9	3	17.6	32	29.1	146	33.0	221	31.8
Caregiver characteristics
Gender
Male	2			12.5	20	18.3	4	23.5	29	26.4	90	20.4	145	20.9
Female	14			87.5	89	81.7	13	76.5	81	73.6	352	79.6	549	79.1
Age, years; mean (SD)	56.6 (12.2)				52.5 (11.3)		54.6 (10.4)		54.8 (12.2)		55.4 (12.8)		54.8 (12.4)
Marital status
Ever married/not single	16			100.0	103	94.5	16	94.1	93	84.5	397	89.8	625	90.1
Unmarried/single	0			0.0	6	5.5	1	5.9	17	15.5	45	10.2	69	9.9
Educational attainment
≥High school	5			38.5	56	52.3	4	25.0	61	56.5	215	48.9	341	49.9
<High school	8			61.5	51	47.7	12	75.0	47	43.5	225	51.1	343	50.1
Relationship characteristics
Caregiver relationship with the patient
Spouse	13			81.3	89	81.7	15	88.2	71	64.5	322	72.9	510	73.5
Adult child	1			6.3	10	9.2	1	5.9	25	22.7	67	15.2	104	15.0
Other	2			12.5	10	9.2	1	5.9	14	12.7	53	12.0	80	11.5
Relationship-process factors
Patient-reported physician prognostic disclosure
Yes	1			6.3	78	71.6	2	11.8	78	70.9	129	29.4	288	41.7
No	15			93.8	31	28.4	15	88.2	32	29.1	310	70.6	403	58.3
Caregiver-reported physician prognostic disclosure
Yes	2			12.5	86	78.9	1	5.9	93	84.5	254	57.6	436	62.9
No	14			87.5	23	21.1	16	94.1	17	15.5	187	42.4	257	37.1
Patient-reported patient-caregiver end-of-life discussion; mean (SD)	9.9 (1.9)				12.4 (3.1)		9.9 (2.7)		12.0 (4.0)		10.7 (3.5)		11.1 (3.6)
Caregiver-reported patient-caregiver end-of-life discussion; mean (SD)	9.4 (3.1)				11.8 (4.1)		9.7 (2.7)		12.9 (4.6)		11.2 (3.8)		11.5 (4.0)
Patient-perceived social support; mean (SD)	61.9 (7.7)				61.5 (9.8)		67.3 (6.2)		65.6 (10.0)		64.1 (10.6)		64.0 (10.3)
Caregiver-perceived social support; mean (SD)	56.5 (8.7)				54.6 (11.9)		63.5 (11.0)		59.2 (11.3)		60.3 (11.5)		59.2 (11.7)

The total number of cases for each characteristic may not equal the sample size because of missing data. Participant and relationship-process characteristics are baseline values. For monthly rows, state percentages are calculated among dyads observed at each time point; the Total % column indicates the proportion of the full study sample with an observed dyadic state.

Factors associated with distinct dyadic death-preparedness states

Few individual or relationship characteristics were associated with distinct dyadic death-preparedness states relative to the discordant state (Table 2). Increasing patient age was associated with a lower likelihood of the cognitive-concordant state (AOR, 0.957; 95% CI, 0.922-0.993) but a higher likelihood of the sufficient-concordant state (AOR, 1.073; 95% CI, 1.034-1.114). Higher patient educational attainment was associated with a greater likelihood of the sufficient-concordant state (AOR, 2.309; 95% CI, 1.304-4.088).

Table 2.

Factors associated with distinct dyadic death-preparedness states.

Dyadic death-preparedness states	Unprepared-concordant				Cognitive-concordant				Emotional-concordant				Sufficient-concordant
Variable	AOR	95% CI		P	AOR	95% CI		P	AOR	95% CI		P	AOR	95% CI		P
Individual-level factors
Patient socio-demographics and disease characteristics
Gender
Male	2.691	0.180	40.292	.473	0.856	0.394	1.864	.696	1.200	0.109	13.165	.881	0.824	0.397	1.708	.602
Female	ref				ref				ref				ref
Age, years	0.963	0.877	1.059	.436	0.957	0.922	0.993	.021	1.005	0.888	1.139	.931	1.073	1.034	1.114	<.001
Marital status
Ever married/not single	3.477	0.051	239.028	.564	1.553	0.549	4.389	.406	1.649	0.059	46.034	.768	1.551	0.584	4.119	.378
Unmarried/single	ref				ref				ref				ref
Educational attainment
>Junior high school	0.915	0.240	3.492	.896	1.479	0.866	2.523	.152	0.976	0.255	3.740	.972	2.309	1.304	4.088	.004
≤Junior High School	ref				ref				ref				ref
Time since diagnosis (months)	0.930	0.848	1.020	.123	0.998	0.986	1.009	.691	0.957	0.893	1.025	.209	1.000	0.989	1.012	.977
Symptom distress	0.987	0.902	1.080	.770	0.985	0.955	1.017	.361	0.933	0.839	1.039	.206	0.985	0.953	1.018	.378
Functional dependence	0.998	0.910	1.095	.966	0.990	0.957	1.024	.564	0.977	0.884	1.079	.642	0.982	0.949	1.017	.311
Co-occurrence of other chronic disease
Yes	1.125	0.622	2.037	.696	1.029	0.810	1.306	.818	1.153	0.638	2.084	.637	1.195	0.942	1.515	.142
No	ref				ref				ref				ref
Caregiver characteristics
Gender
Male	1.569	0.099	24.839	.749	1.132	0.517	2.477	.756	1.711	0.144	20.258	.670	1.034	0.486	2.199	.931
Female	ref				ref				ref				ref
Age, years	1.005	0.916	1.102	.918	0.999	0.966	1.034	.966	0.946	0.839	1.067	.365	0.985	0.950	1.020	.391
Marital status
Ever married/not single	NE	NE	NE	NE	0.829	0.336	2.045	.684	2.363	0.051	110.539	.661	1.106	0.459	2.665	.822
Unmarried/single	ref				ref				ref				ref
Educational attainment
≥High school	0.963	0.232	3.986	.958	0.788	0.454	1.367	.396	0.414	0.090	1.894	.255	0.945	0.518	1.725	.854
<High school	ref				ref				ref				ref
Relationship characteristics
Caregiver relationship with the patient
Spouse	0.262	0.004	16.638	.527	1.397	0.424	4.599	.583	1.750	0.033	93.542	.783	0.492	0.164	1.481	.207
Adult child	0.372	0.003	52.228	.695	1.401	0.265	7.409	.691	0.460	0.003	79.357	.770	1.149	0.312	4.235	.835
Other	ref				ref				ref				ref
Relationship-process factors
Patient-reported physician prognostic disclosure
Yes	0.191	0.022	1.674	.135	3.416	2.193	5.323	<.001	0.231	0.035	1.534	.129	3.743	2.375	5.899	<.001
No	ref				ref				ref				ref
Caregiver-reported physician prognostic disclosure
Yes	0.134	0.032	0.563	.006	2.411	1.470	3.957	.001	0.065	0.012	0.346	.001	2.854	1.704	4.779	<.001
No	ref				ref				ref				ref
Patient-reported patient-caregiver end-of-life discussion	0.956	0.768	1.189	.684	1.190	1.106	1.279	<.001	0.800	0.650	0.985	.035	1.040	0.972	1.113	.257
Caregiver-reported patient-caregiver end-of-life discussion	1.019	0.841	1.235	.847	1.016	0.958	1.077	.600	0.839	0.694	1.015	.070	1.051	0.992	1.115	.092
Patient-perceived social support	0.963	0.898	1.033	.291	0.956	0.929	0.974	<.001	1.059	0.988	1.135	.105	0.998	0.974	1.023	.861
Caregiver-perceived social support	0.967	0.917	1.020	.215	0.945	0.927	0.963	<.001	1.050	0.991	1.113	.096	0.987	0.967	1.007	.193

The discordant state served as the reference outcome. All models were adjusted for time until death; corresponding estimates are presented in Table S2. Estimates for the uncommon unprepared- and emotional-concordant states should be interpreted cautiously because sparse observations may produce unstable estimates and wide CIs. Boldface indicates statistical significance (P < .05).

Abbreviations: AOR, adjusted odds ratio; NE, estimate not obtainable because of sparse/zero cell counts; ref: reference.

In contrast, relationship-process factors more consistently distinguished concordant states from the discordant state. Patient- and caregiver-reported physician prognostic disclosure was associated with higher likelihoods of cognitive and sufficient concordance (patient-reported disclosure: AOR, 3.416; 95% CI, 2.193-5.323 and AOR, 3.743; 95% CI, 2.375-5.899, respectively; caregiver-reported disclosure: AOR, 2.411; 95% CI, 1.470-3.957 and AOR, 2.854; 95% CI, 1.704-4.779, respectively). Caregiver-reported physician prognostic disclosure was additionally associated with lower likelihoods of the unprepared- and emotional-concordant states (AOR, 0.134; 95% CI, 0.032-0.563 and AOR, 0.065; 95% CI, 0.012-0.346, respectively). Because the unprepared- and emotional-concordant states were uncommon, these estimates were imprecise and should be interpreted cautiously.

Greater patient-reported—but not caregiver-reported—EOL discussions were associated with a higher likelihood of the cognitive-concordant state (AOR, 1.190; 95% CI, 1.106-1.279) but a lower likelihood of the emotional-concordant state (AOR, 0.800; 95% CI, 0.650-0.985).

Higher perceived social support reported by both patients (AOR, 0.956; 95% CI, 0.929-0.974) and caregivers (AOR, 0.945; 95% CI, 0.927-0.963) was associated with a lower likelihood of the cognitive-concordant state.

Discussion

This study identified factors associated with distinct dyadic death-preparedness states among patients with advanced cancer and their family caregivers. Relationship-process factors more consistently distinguished concordant states from the discordant state, whereas individual and relationship characteristics played comparatively limited roles. Together, these findings reinforce the importance of conceptualizing death preparedness as a dyadic phenomenon and of considering prognostic communication and relational context when assessing how patients and caregivers prepare for death as it approaches.

Patient age and educational attainment were the only individual characteristics associated with dyadic death-preparedness states. Increasing patient age was associated with a lower likelihood of cognitive concordance but a higher likelihood of sufficient concordance. In Taiwanese/Chinese contexts, family-consent practices may result in withholding prognostic information from older patients to protect them from distress,³¹ potentially reducing symmetrical prognostic awareness within dyads. Older age may simultaneously be associated with greater acceptance of illness and mortality,^32–34 which may partly account for its association with sufficient concordance. Higher patient educational attainment was also associated with sufficient concordance, possibly reflecting greater resources for understanding prognostic information and preparing jointly for approaching death.^20,35,36 Notably, disease-related characteristics, including symptom distress, functional dependence, comorbidity, and time since diagnosis, were not associated with dyadic death-preparedness states. Although symptom distress and functional dependence were associated with patients’ distinct death-preparedness states in our previous work,¹⁷ these clinical characteristics may be less closely related to dyadic concordance, which requires alignment between patients’ and caregivers’ cognitive and emotional preparedness as death approaches.

Physician prognostic disclosure reported by patients and caregivers was associated with higher likelihoods of the cognitive- and sufficient-concordant states. Prognostic disclosure has been linked to more accurate prognostic awareness among patients with cancer²⁰ and caregivers^33,37 and to greater emotional acceptance of terminal illness or death in both groups.^{32,33,38–40} Our previous studies further showed that disclosure was associated with higher likelihoods that patients and caregivers achieved cognitive and sufficient preparedness individually. The present findings extend this evidence by showing that participant-reported disclosure is also associated with preparedness shared within the patient-caregiver dyad. Caregiver-reported physician prognostic disclosure was additionally associated with lower likelihoods of the unprepared- and emotional-concordant states. Under Taiwanese family-consent practices,³¹ prognostic information may be disclosed to caregivers before patients; because both states are characterized by limited prognostic awareness, this pattern may partly explain these associations. However, because these states were uncommon and the corresponding estimates were imprecise, these findings warrant cautious interpretation and require replication. Overall, the more robust findings for cognitive and sufficient concordance suggest that whether patients and caregivers report having received prognostic disclosure may be associated with whether they subsequently share an understanding of approaching death.

Greater patient-reported, but not caregiver-reported, EOL discussions were associated with a higher likelihood of cognitive concordance and a lower likelihood of emotional concordance. This pattern suggests that greater patient-reported EOL discussion may relate differently to the cognitive and emotional components of dyadic preparedness, being associated with higher cognitive concordance but lower emotional-only concordance. Mutual acknowledgement and understanding may help patients and caregivers construct a shared view of illness and care,⁴¹ whereas their emotional responses to approaching death may remain different. One possible explanation is that caregivers may become better prepared for deterioration, whereas patients may experience greater anticipatory grief as approaching death becomes more salient. Thus, EOL discussion may be associated with cognitive alignment without necessarily being accompanied by parallel emotional adaptation within the dyad.

The absence of associations for caregiver-reported discussions further suggests that the occurrence of communication alone may not capture the aspects of EOL communication associated with dyadic preparedness. Patients and caregivers may differ in how they perceive or experience the same discussions, and caregivers may also limit or filter communication in an effort to protect the patient from distress, a pattern often described as protective buffering.⁴² Although these mechanisms require further investigation, the divergence between patient- and caregiver-reported associations suggests that both perspectives may be important when evaluating EOL communication.

Higher perceived social support reported by both patients and caregivers was associated with a lower likelihood of cognitive concordance. Although social support has not been consistently identified as a correlate of prognostic awareness,²⁰ greater family contact has been associated with lower patient awareness.⁴³ Within Taiwanese family-consent practices,³¹ strong supportive involvement may coexist with family efforts to manage or withhold prognostic information to protect patients from distress. Social support should therefore not be interpreted as inherently impeding shared preparedness; rather, its association with preparedness may depend on how supportive networks are involved in direct prognostic communication. Given the broader social-network focus of the MOS-SSS, these mechanisms remain speculative.

However, these findings should also be interpreted within Taiwan’s sociocultural context. Family involvement in serious-illness and EOL decision-making is often substantial in East Asian settings, and filial responsibility may shape how patients and caregivers respond to incurable illness and approaching death.^31,34,44,45 These cultural and relational norms may influence how shared preparedness is expressed within patient-caregiver dyads and should be considered when interpreting the present associations.

Limitations

This study has several limitations. First, generalizability may be limited by recruitment from a single medical center in Taiwan, and the extended recruitment period may have introduced secular changes in clinical practice. Because attitudes toward death, treatment limitation, and family involvement vary both across and within societies according to cultural, ethnic, religious, and health-system contexts, including trust in medical institutions, findings from this Taiwanese cohort may not generalize to populations with different sociocultural and health-care contexts. Additionally, because the dyadic design required participation of a family caregiver, patients without an available or participating family caregiver were not represented. Such patients may experience different processes of prognostic understanding, emotional acceptance, communication, and support as death approaches; therefore, our findings regarding dyadic death preparedness should not be generalized to patients without family caregiver involvement.

Second, despite the large sample, the unprepared- and emotional-concordant states were uncommon, partly because dyadic states were derived by cross-classifying patient and caregiver preparedness states and because their prevalence changed as death approached.⁸ Sparse observations produced wide CIs and some non-estimable or potentially unstable estimates; findings involving these states should therefore be interpreted cautiously and should not be overinterpreted clinically.

Third, patient and caregiver death-preparedness states were assigned using highest posterior probabilities from previously identified latent states,⁸ and dyadic states were subsequently derived by cross-classification; emotional preparedness was also dichotomized at the sample median. These approaches facilitated clinically interpretable state classification but may have introduced misclassification, and latent classification uncertainty was not incorporated into the present analyses.

Fourth, although lagging time-varying predictors strengthened temporal ordering, the observational design precludes causal inference. Moreover, physician prognostic disclosure and accurate prognostic awareness are conceptually related, and accurate prognostic awareness contributed to defining the death-preparedness states; thus, associations between prior prognostic disclosure and subsequent state membership may partly reflect this conceptual relationship. Lagging reduced contemporaneous measurement overlap but did not eliminate this relatedness. The analytic approach also did not evaluate transitions between dyadic death-preparedness states.

Finally, prognostic disclosure and EOL discussions did not capture communication quality. Physician prognostic disclosure was based on participant report rather than verified clinician documentation and may therefore have been affected by recall or differences in perception of prognostic communication. Unmeasured relational factors, including relationship quality, protective buffering, and dyadic coping, may also have influenced the findings.

Conclusions and clinical implications

Distinct dyadic death-preparedness states were more consistently associated with relationship-process factors than with individual or relationship characteristics. Prognostic disclosure, EOL discussions, and perceived social support were associated with distinct patterns of shared cognitive and emotional preparedness for death. These findings underscore the importance of considering both members of the patient-caregiver dyad when assessing preparedness for approaching death and of attending to prognostic communication and relational support as correlates of shared preparedness.

For oncology practice, assessment should distinguish prognostic understanding from emotional preparedness in both patients and caregivers rather than assuming that shared understanding necessarily indicates shared emotional readiness. Clinicians may also need to consider who has received and understood prognostic information and elicit both patients’ and caregivers’ perspectives on EOL discussions, rather than relying solely on whether communication has occurred.^46,47 Because greater perceived social support was not uniformly associated with more concordant preparedness, clinicians should also explore how family and broader social networks are involved in prognostic communication rather than assuming that greater support necessarily corresponds to shared preparedness. These relational processes may provide clinically relevant context for understanding how patient-caregiver dyads prepare for approaching death.

Author contributions

Fur-Hsing Wen (Conceptualization, Formal analysis, Investigation, Writing—original draft, Writing—review & editing), Chia-Hsun Hsieh (Conceptualization, Data curation, Investigation, Writing—original draft, Writing—review & editing), Wen-Chi Chou (Conceptualization, Data curation, Investigation, Writing—original draft, Writing—review & editing), Jen-Shi Chen (Conceptualization, Data curation, Investigation, Writing—original draft, Writing—review & editing), Wen-Cheng Chang (Conceptualization, Data curation, Investigation, Writing—original draft, Writing—review & editing), and Siew Tzuh Tang (Conceptualization, Data curation, Funding acquisition, Investigation, Project administration, Resources, Supervision, Writing—original draft, Writing—review & editing)

Supplementary material

Supplementary material is available at The Oncologist online.

Funding

This study was supported by the Ministry of Science and Technology (MOST 104-2314-B-182-027-MY3 and MOST 105-2314-B-182-040-MY3), National Health Research Institutes (NHRI-EX111-10704PI), and Chang Gung Memorial Hospital (BMRP888). The funders had no role in the design or conduct of the study; collection, management, analysis, or interpretation of the data; or preparation, review, or approval of the manuscript.

Conflicts of interest

None declared.

Data availability

The sharing of anonymized data from this study is restricted because of ethical and legal requirements. The data contain sensitive personal health information protected under Taiwan’s Personal Data Protection Act, and data sharing is therefore subject to Institutional Review Board (IRB) approval. Per the Chang Gung Memorial Hospital IRB, data supporting the findings of this study are restricted to the primary investigative team. Requests for anonymized data from investigators outside the team will be reviewed by the research team and submitted to the Chang Gung Medical Foundation IRB for approval. Upon IRB approval, the data may be available from the corresponding author (Siew Tzuh Tang) upon reasonable request.

Ethics approval

The Chang Gung Medical Foundation Institutional Review Board approved both studies (approval numbers 103-7015B and 104-9388B). Study procedures were conducted in accordance with the ethical standards of the institutional research committee and the Declaration of Helsinki.

Consent to participate

Written informed consent was obtained from all participants.