Association between family-perceived quality of death and dying, prognostic awareness, and end-of-life discussions in five common diseases: a nationwide mortality follow-back study Authors: Yoko Nakazawa, Mitsunori Miyashita, Tatsuya Morita, Masanori Mori, Yoshiyuki Kizawa, Yasuyuki Okumura, Shohei Kawagoe, Hiroshi Yamamoto, Emi Takeuchi, Risa Yamazaki, Asao Ogawa Published: 2026 ABSTRACT AbstractIntroductionDiscussing prognosis and end-of-life (EOL) care with patients and physicians is a sensitive and critical aspect of EOL care. However, the association between prognostic awareness and EOL discussions and the quality of death and dying (QOD) remains uncertain. This study aimed to describe prognostic awareness and participation in EOL discussions among patients with common causes of death and to clarify their association with QOD.MethodsRetrospective nationwide mortality follow-back survey in Japan (Feb 2019; Feb 2020). Bereaved families of patients who died in hospitals or at home from cancer, heart disease, cerebrovascular disease, pneumonia, or renal failure. The primary outcome was family-reported QOD. Exposures of interest included family-reported prognostic awareness and EOL discussions, including care settings and code statuses. Analyses were conducted separately by disease category and place of death. Logistic regression and generalized linear models were used for analysis.ResultsOf the 115,861 family members who were sent the questionnaires, we finally analyzed 50,641 responses. Accurate patient prognostic awareness ranged from 56.8% to 70.2% for patients with cancer and from 30.5% to 53.3% for patients with non-cancer diseases. Patient-physician discussions about care settings occurred in 36.4%–54.8% of hospital deaths and 67.8%–81.5% of home deaths across diseases. There were small differences in the QOD scores between patients with accurate and inaccurate prognostic awareness across most subgroups. However, participation in EOL discussions was associated with higher QOD scores than non-participation. The largest QOD score difference was observed in patients who died from pneumonia in hospitals, with discussions about care settings (64.0 versus without 42.6) showing a difference of 21.4 (95% confidence interval: 16.3, 26.5).ConclusionsParticipation in EOL discussions was associated with higher family-reported QOD across diseases and places of death in this observational study. Similar patterns were observed among patients who died in hospital settings. These findings suggest that EOL discussions may be related to QOD beyond the final place of death.Supplementary InformationThe online version contains supplementary material available at https://doi.org/10.1186/s12904-026-02205-x.Keywords: Quality of life, Terminal Care, Prognosis, End of Life, Communication Introduction Discussions regarding prognosis and end-of-life (EOL) care between patients and physicians are critical components of EOL care [1–3]. Early engagement in EOL discussions has been associated with a greater likelihood that patients with advanced cancer receive their preferred care [4, 5], thereby reducing aggressive medical interventions that are related to poor quality of life [4, 6–8]. In particular, accurate prognostic awareness has been associated with greater concordance between patients’ goals of care and the care received, as well as improved quality of life among patients with cancer [5, 9, 10]. However, most of the studies on prognostic awareness and EOL discussions have been conducted in cancer patients, and less is known for other chronic, life-limiting illnesses [9, 11–13]. While prognostic awareness reflects patients’ understanding of their illness status, EOL discussions involve communication processes through which care preferences are shared. In Japan, studies of EOL discussions have primarily focused on patients with cancer [6, 10, 14], and their dissemination remains limited. Japan has a family-centered culture, and patients often prefer that the family be informed of the prognosis first [15]. In one study, the majority of patients with advanced cancer had a prognostic awareness that was more optimistic than their actual survival time [14]. Beyond cultural factors, characteristics of the Japanese health care system also shape the context of EOL communication. Under the universal health insurance system, access to specialist and hospital care is relatively open, and primary care physicians generally do not function as strict gatekeepers. As a result, end-of-life care is provided across multiple settings, including hospitals, home-based care supported by community physicians and visiting nurse services, and palliative care units (PCUs). Against this background, national policy initiatives are underway to strengthen EOL decision-making and promote shared decision-making. These initiatives include efforts to promote advanced care planning for people with serious illness, as well as cancer-specific strategies such as the Basic Plan to Promote Cancer Control Programs [16, 17]. We therefore examined how prognostic awareness and participation in EOL discussions were associated with quality of death and dying (QOD) across diseases and care settings. In this study, QOD refers to the overall quality of the end-of-life experience as perceived by bereaved family members. The construct has been assessed using measures such as the Quality of Dying and Death (QODD) developed by Curtis et al. [18]. In the present study, QOD was assessed using the validated Japanese Good Death Inventory (GDI) [19]. This study aimed to describe patients’ prognostic awareness and participation in EOL discussions and to clarify their associations with the QOD based on a nationwide survey of bereaved families of patients who died from cancer, heart disease, cerebrovascular disease, pneumonia, or renal failure. Methods Design and participants This observational study was a secondary analysis of data from a nationwide mortality follow-back survey conducted between February 2019 and February 2020. The original study aimed to examine family-reported outcomes of EOL care by place and cause of death among patients with five common causes of death (cancer, heart disease, cerebrovascular disease, pneumonia, and renal failure) in Japan [20]. The survey used the most recent mortality data available from the 2017 and 2018 surveys. The questionnaire was mailed to bereaved families to inquire about their EOL experiences 13–25 months after the patients’ deaths. Each questionnaire was addressed to the decedents’ family using the decedents’ names obtained from death certificate records. We requested that an adult family member who had been the primary caregiver complete the survey. Inclusion criteria were as follows: (1) age at death ≥ 20 years and (2) death in one of three places: hospital, long-term care facility, or at home. The target population was randomized according to the place of death for each of the five common causes of death in Japan [20, 21]. In this study, deaths in long-term care facilities were excluded to focus on hospital and home deaths, which account for more than 90% of deaths in Japan. This exclusion was also justified because care processes and decision-making contexts in long-term care facilities differ from those in hospital and home settings [22]. In addition, deaths occurring within one month of diagnosis were excluded as a proxy for sudden or early deaths, as such cases were unlikely to have provided sufficient opportunities for prognostic awareness or end-of-life discussions. The 13–25 month interval after death represents the earliest feasible period for conducting a mortality follow-back survey under Japan’s secondary use regulations for national vital statistics data. The detailed process of case selection and exclusions is illustrated in Fig. 1. Flow diagram of case selection. a The five selected primary causes of death were cancer, heart disease, cerebrovascular disease, pneumonia, and renal failure. Cancer-related deaths included those occurring in both 2017 and 2018, whereas other causes were restricted to deaths occurring in 2017. Outcome measurements All study variables were based on reports from bereaved family members rather than direct responses from the patients themselves. Quality of death and dying The QOD was assessed using the Japanese version of the Good Death Inventory (GDI), which includes 10 core domains and has been shown to have adequate reliability and validity [19]. Family members were asked to rate the decedent’s experiences during the last month of life on a 7-point Likert scale (ranging from 1: absolutely disagree to 7: absolutely agree) and “unknown.” We converted scores from 70 to 100 points using proportional scaling, with no minimum score, resulting in a range of 14.3 to 100, with a higher score indicating a better outcome. For ease of interpretation, scores were linearly rescaled as (raw total score/70) × 100. The list of GDI items is provided in Supplementary Table 1. Primary exposure of interest Prognostic awareness We modified and used a scale adapted from an existing study to assess prognostic awareness [4]. Family members were asked: “How did the patient perceive his/her illness one month before death?” Response options included “unaware of illness,” “not seriously ill,” “seriously ill but not terminal,” “seriously ill and terminal,” and “unknown.” Consistent with prior studies, responses were dichotomized to distinguish clear terminal awareness from other states for analytic clarity [4]. We considered responses indicating “seriously ill and terminal” to be indicative of accurate prognostic awareness. EOL discussions EOL discussions were defined as discussions with physicians about the preferred place of care at the EOL or resuscitation [6, 23]. Family members were asked: “Had the patient discussed with the attending physician where he/she wished to spend the end of life by one month before death?” Additional items assessed whether the patient and physician had discussed the patient’s code status and whether the family and physician had discussed code status. The code status indicated patients’ preferences regarding medical intervention in life-threatening situations, including directives such as Do Not Resuscitate (DNR) and Full Code. Specifically, the patient’s wishes regarding resuscitation efforts in the event of cardiac or respiratory arrest. The response options were “absolutely disagree,” “disagree,” “agree,” “absolutely agree,” “not applicable,” and “unknown.” For analytic purposes, responses of “agree” and “absolutely agree” were grouped to indicate that a discussion had occurred. Statistical analysis For each cause of death, we analyzed the outcomes by subgroups according to the place of death. In Japan, only a subset of patients with cancer receive specialized palliative care in palliative care units (PCUs); therefore, we extracted PCU deaths from hospital deaths and analyzed them separately. Absolute differences and odds ratios (ORs) were calculated for the proportion of prognostic perceptions and EOL discussions according to the place of death. To assess prognostic perceptions, EOL discussions, and their association with QOD, absolute differences in GDI scores were compared between the groups. Overall, 18.9% of respondents had at least one missing GDI item. Therefore, missing item responses within the GDI were imputed using the individual mean of the non-missing items for each respondent (i.e., person-mean imputation), following consultation with the scale developer. To assess the potential impact of the imputation on the estimates, we conducted a sensitivity analysis by repeating the main analyses using complete-case data for each outcome; the conclusions were consistent with the primary analysis. The GDI domain scores were calculated based on the disease. The proportions of prognostic awareness and EOL discussion participation were analyzed using logistic regression models, and comparisons of GDI scores were analyzed using generalized linear models. The model covariates included patient sex, age, and time since diagnosis. Because our research question was primarily descriptive and exploratory rather than causal, we did not attempt to construct a fully specified causal model or adjust for all potential confounders within a directed acyclic graph framework. Instead, we selected a limited set of a priori covariates—patient age, sex, and time since diagnosis—that were clinically relevant, consistently available across all disease groups, and conceptually pre-exposure variables. Comorbidities and access to palliative care were not available from bereaved family reports and, therefore, could not be included in the models. All analyses were performed using SAS software (version 9.4; SAS Institute, Cary, NC, USA). Results Sample characteristics In the original study, 62,576 valid responses were received from the 115,861 family members who received the questionnaires (Fig. 1). This study analyzed 50,641 responses, excluding responses from family members of sudden-death patients. Table 1 shows the numbers and proportions of sociodemographic characteristics by place of death for each disease. Table 1Characteristics of decedents and respondents by place of death by five diseasesSample No.CancerHeart diseaseCVDPneumoniaRenal failureHospitalHomePCUP valueHospitalHomeP valueHospitalHomeP valueHospitalHomeP valueHospitalHomeP value2227717681677512681034239202230261290384DecedentsAge, mean(SD),ya77.3(11.3)77.6(11.5)75.0(11.2)<.000185.9(9.5)83.5(10.9)<.000183.3(10.5)83.2(12.4)0.9286.5(8.6)86.4(9.8)0.9084.9(8.7)87.2(8.6)<0.001 20s to 50s1502(6.7)1136(6.4)585(8.6)<0.00125(2.0)35(3.4)<0.00111(4.6)10(5.0)0.582(0.9)5(1.9)0.334(1.4)5(1.3)0.03 60s3711(16.7)2764(15.6)1355(20.0)66(5.2)88(8.5)13(5.4)17(8.4)10(4.4)9(3.5)14(4.8)11(2.9) 70s6460(29.0)5161(29.2)2250(33.2)141(11.1)178(17.2)44(18.4)28(13.9)20(8.7)35(13.4)38(13.1)42(10.9) 80s7850(35.2)6157(34.8)2097(31.0)533(42.0)399(38.6)97(40.6)82(40.6)107(46.5)106(40.6)141(48.6)156(40.6) 90s and above2754(12.4)2463(13.9)488(7.2)503(39.7)334(32.3)74(31.0)65(32.2)91(39.6)106(40.6)93(32.1)170(44.3)Sex, Femalea8588(38.6)7080(40.0)2871(42.4)<0.001716(56.5)532(51.5)0.02108(45.2)99(49.0)0.42104(45.2)118(45.2)1.00150(51.7)166(43.2)0.03Time since Diagnosis ≤ 1 Year11100(51.2)8125(47.1)3173(47.4)<0.001429(40.4)148(23.9)<0.001115(53.0)45(31.5)<0.001136(66.7)90(46.9)<0.00196(36.9)104(32.2)0.42 ≤ 5 Years7913(36.5)6828(39.6)2732(40.8)288(27.1)189(30.5)58(26.7)43(30.1)51(25.0)40(20.8)65(25.0)93(28.8) > 5 Years2669(12.3)2311(13.4)788(11.8)344(32.4)283(45.7)44(20.3)55(38.5)17(8.3)62(32.3)99(38.1)126(39.0)Impaired communication2800(12.7)1271(7.3)657(9.8)<0.001367(29.5)86(8.5)<0.001145(63.0)51(25.5)<0.001101(45.1)67(26.3)<0.00172(25.4)52(13.7)0.00Coexistence of Dementia2625(12.2)1973(11.5)573(8.7)<0.001394(33.3)189(19.4)<0.00188(40.2)55(30.1)0.03101(46.5)101(41.1)0.23185(33.5)250(32.1)0.71Preferred Care Setting Hospital3853(26.6)140(0.9)663(13.4)<0.001125(17.0)4(0.6)<0.00125(21.2)0(0.0)<0.00124(20.2)4(2.0)<0.00135(18.9)3(0.9)<0.001 Long Term Care213(1.5)124(0.8)30(0.6)40(5.5)10(1.4)8(6.8)1(0.7)8(6.7)5(2.5)6(3.2)11(3.4) Home9590(66.3)15357(97.4)2387(48.2)562(76.6)689(98.0)83(70.3)140(99.3)87(73.1)194(95.6)142(76.8)311(95.1) Palliative Care Unit811(5.6)152(1.0)1876(37.9)7(1.0)0(0.0)2(1.7)0(0.0)0(0.0)0(0.0)2(1.1)2(0.6)Code status, DNR12203(94.3)10736(97.2)4351(97.0)<0.001521(84.0)330(81.9)0.37105(91.3)91(92.9)0.68110(92.4)140(97.2)0.07147(90.7)217(96.0)0.03Respondents Age, mean(SD)64.6(11.9)65.5(11.9)64.3(12.1)<.000165.8(10.6)64.7(11.6)0.0267.7(10.9)65.7(10.5)0.0565.9(10.1)67.9(11.0)0.0563.9(11.5)65.9(10.4)0.02 Sex, Female13988(63.6)12334(70.5)4281(63.6)<0.001697(55.7)591(58.3)0.22141(60.8)130(64.7)0.40121(54.3)162(63.8)0.03176(60.9)265(70.1)0.01 Relationship to Decedent Spouse9843(44.8)8658(49.5)3351(49.8)<0.001266(21.3)226(22.4)0.5580(34.6)55(27.6)0.4245(20.2)74(29.0)0.0156(19.4)97(25.6)0.31 Child8723(39.7)6494(37.2)2425(36.0)691(55.4)559(55.5)111(48.1)101(50.8)138(61.9)138(54.1)178(61.8)205(54.1) Child-in-law1756(8.0)1401(8.0)437(6.5)173(13.9)124(12.3)24(10.4)22(11.1)27(12.1)24(9.4)35(12.2)51(13.5) Parent504(2.3)349(2.0)147(2.2)37(3.0)23(2.3)7(3.0)7(3.5)1(0.5)10(3.9)7(2.4)8(2.1) Other1145(5.2)574(3.3)372(5.5)81(6.5)75(7.5)9(3.9)14(7.0)12(5.4)9(3.5)12(4.2)18(4.8)Abbreviations: PCU Palliative care unit, CVD Cerebrovascular disease, DNR Do-not-resuscitate ordera Data were calculated using death certificate data of vital statistics Characteristics of decedents and respondents by place of death by five diseases Abbreviations: PCU Palliative care unit, CVD Cerebrovascular disease, DNR Do-not-resuscitate order a Data were calculated using death certificate data of vital statistics The mean age of patients with cancer was from 75.0 to 77.6 years, which was younger than that of patients with other diseases, ranging from 83.2 to 87.2 years. Regardless of diagnosis, the preferred care setting for patients was primarily at home, with percentages ranging from 48.2% to 99.3%. There was a high proportion of DNR orders, ranging from 81.9% to 97.2%, across all diagnoses. Prognostic awareness and EOL discussions Table 2 shows the proportions of accurate patient prognostic awareness and participation in EOL discussions as reported by the family. Family-reported accurate prognostic awareness was highest in the PCUs for patients with cancer, ranging from 56.8% (95% CI 56.1 to 57.5) in the in-hospital group to 70.2% (95% CI 69.0 to 71.3) in the PCU group. Table 2Proportions of prognostic awareness and EOL discussions by place of death for five diseasesCrude proportions (95% CI)Hospita (ref.)HomeDifferenceaCrude ORAdjusted OR (95% CI)bPCUDifferenceaCrude ORAdjusted OR (95% CI)bCancer Prognostic Awareness56.8(56.1, 57.5)68.6(67.8, 69.3)11.8(10.7, 12.8)1.661.71(1.63, 1.79)70.2(69.0, 71.3)13.3(12.0, 14.7)1.791.70(1.59, 1.81) Patient-Physician Discussion on Care Setting54.8(53.9, 55.6)81.5(80.9, 82.2)26.8(25.7, 27.9)3.653.62(3.43, 3.83)66.8(65.4, 68.1)12.0(10.4, 13.6)1.661.61(1.50, 1.73) Patient-Physician Discussion on Code Status.60.9(60.0, 61.7)67.3(66.4, 68.3)6.5(5.2, 7.7)1.331.31(1.24, 1.38)71.9(70.5, 73.3)11.0(9.4, 12.7)1.641.62(1.50, 1.76) Patient-family Discussion on End-of-life 48.3(47.6, 49.1)64.0(63.2, 64.8)15.7(14.6, 16.7)1.901.88(1.80, 1.97)59.5(58.3, 60.8)11.2(9.7, 12.6)1.571.56(1.47, 1.66)Heart disease Prognostic Awareness38.0(34.7, 41.2)33.9(30.3, 37.5)-4.1(-8.9, 0.8)0.840.97(0.76, 1.24)- Patient-Physician Discussion on Care Setting42.7(38.4, 47.1)75.7(71.1, 80.4)33.0(26.7, 39.3)4.184.50(3.18, 6.43)- Patient-Physician Discussion on Code Status.54.3(50.3, 58.3)56.5(50.9, 62.1)2.2(-4.6, 9.0)1.091.03(0.75, 1.42)- Patient-family Discussion on End-of-life 41.9(38.7, 45.0)42.6(38.7, 46.4)0.7(-4.3, 5.6)1.031.30(1.02, 1.65)-CVD Prognostic Awareness37.5(28.7, 46.3)30.5(22.1, 38.9)-7.0(-19.0, 5.0)0.730.67(0.35, 1.27)- Patient-Physician Discussion on Care Setting36.6(27.1, 46.2)67.8(57.9, 77.6)31.1(17.7, 44.6)3.643.36(1.74, 6.65)- Patient-Physician Discussion on Code Status.53.4(44.3, 62.5)67.5(57.0, 78.0)14.1(0.5, 27.8)1.811.76(0.91, 3.48)- Patient-family Discussion on End-of-life 38.5(31.6, 45.4)55.6(47.1, 64.2)17.2(6.3, 28.0)2.012.12(1.27, 3.60)-Pneumonia Prognostic Awareness32.3(24.0, 40.5)37.7(30.1, 45.2)5.4(-5.7, 16.4)1.271.29(0.73, 2.30)- Patient-Physician Discussion on Care Setting36.4(27.2, 45.7)69.2(61.9, 76.4)32.7(21.1, 44.3)3.913.96(2.21, 7.20)- Patient-Physician Discussion on Code Status.60.7(51.5, 69.9)59.7(51.3, 68.1)-1.0(-13.3, 11.3)0.960.93(0.51, 1.69)- Patient-family Discussion on End-of-life 39.1(32.0, 46.2)56.8(49.7, 63.8)17.6(7.7, 27.6)2.042.62(1.62, 4.29)-Renal failure Prognostic Awareness52.1(45.4, 58.8)53.3(47.3, 59.2)1.2(-7.7, 10.1)1.051.23(0.83, 1.83)- Patient-Physician Discussion on Care Setting41.1(33.4, 48.9)80.3(75.4, 85.2)39.2(30.1, 48.2)5.845.71(3.54, 9.38)- Patient-Physician Discussion on Code Status.58.8(50.9, 66.7)69.2(62.7, 75.7)10.4(0.3, 20.5)1.571.49(0.91, 2.45)- Patient-family Discussion on End-of-life 46.8(40.4, 53.2)60.9(55.4, 66.5)14.1(5.7, 22.5)1.772.05(1.41, 3.01)-Abbreviations: EOD End-of-Life, ref. Reference, PCU Palliative care unit, CVD cerebrovascular diseaseaDifference indicates the absolute differences between each group and the reference hospitalbAdjusted OR: Odds ratio calculated using logistic regression, adjusted for sex, age, and time since diagnosis Proportions of prognostic awareness and EOL discussions by place of death for five diseases Abbreviations: EOD End-of-Life, ref. Reference, PCU Palliative care unit, CVD cerebrovascular disease aDifference indicates the absolute differences between each group and the reference hospital bAdjusted OR: Odds ratio calculated using logistic regression, adjusted for sex, age, and time since diagnosis In contrast, prognostic awareness was substantially lower among patients with non-cancer diagnoses. The proportions ranged from 30.5% to 53.3% across heart disease, cerebrovascular disease, pneumonia, and renal failure, indicating consistently lower levels of awareness compared to cancer across care settings. Family-reported patient–physician discussions about care settings were 36.4% to 54.8% in the in-hospital groups, compared to 67.8% to 81.5% in the home groups. This pattern of higher participation in home settings was observed across both cancer and non-cancer diagnoses (adjusted OR range 3.65 to 5.84). Comparison of quality of death and dying Figure 2 shows a comparison of family-reported QOD scores with prognostic awareness and EOL discussions. The differences in family-reported QOD between patients with accurate and inaccurate prognostic awareness were generally small across most subgroups. The QOD was higher for inaccurate than for accurate prognostic awareness in the home care group for heart diseases (67.8 vs. 71.9; difference − 4.1, 95% CI − 6.4 to − 1.8) and in the in-hospital care group for pneumonia (53.0 vs. 61.1; difference − 8.2, 95% CI − 14.8 to − 1.5). QOD Comparisons with Prognostic Awareness and EOL Discussions by Place of Death for Five Diseases. Abbreviations: QOD, quality of death and dying; EOL, end-of-life; CVD, cerebrovascular disease; P-PD, patient-physician discussion; P-FD, patient-family discussion. The plots show the absolute differences between the absence and presence groups, with bars indicating 95% confidence intervals. a QOD was measured using the Good Death Inventory-core, with scores ranging between 14.3-100, adjusted for sex, age, and time since diagnosis. b Prognostic awareness was compared between the accurate group and the inaccurate group However, comparing family-reported QOD scores based on participation in EOL discussions, the QOD scores were higher for most diseases in the discussion participation group. The largest difference in the QOD was in patient–physician discussions about care settings in the in-hospital group for pneumonia (participation 64.0 (95% CI 56.9 to 71.0) vs. non-participation 42.6 (95% CI 35.5 to 49.7); difference 21.4 (95% CI 16.3 to 26.5)) and for heart disease (participation 64.5 (95% CI 60.2 to 68.9) vs. non-participation 49.8 (95% CI 45.7 to 53.9); difference 14.7 (95% CI 11.8 to 17.7)). Differences of comparable magnitude were also observed for patient–physician discussions of code status. For example, in the in-hospital group for pneumonia, QOD scores were 57.0 (95% CI 49.8 to 64.2) in the participation group and 40.5 (95% CI 32.1 to 48.8) in the non-participation group (difference 16.5, 95% CI 10.8 to 22.2) (Fig. 2). Participation in EOL discussions was associated with higher domain scores across the five QOD domains (Supplementary Table 1). For example, among patients with pneumonia, patient–physician discussions about care setting were associated with higher scores for having some pleasure in daily life (4.2 vs. 2.7; difference 1.5, 95% CI 1.1 to 1.9) and trusting physicians (5.4 vs. 3.8; difference 1.6, 95% CI 1.3 to 2.0). Discussion A key strength of this study is its large nationwide sample of bereaved family responses, which allowed precise comparisons across five common causes of death and place-of-death subgroups. Prognostic awareness and EOL discussions In this study, we clarified family-reported patient prognostic awareness, the actual circumstances of EOL discussions, and examined their association with QOD using place-of-death subgroups for five common causes of death. One of the findings was that a particularly high proportion of cancer deaths occurring at homes and PCUs were reported by family members as cases in which patients recognized their condition as incurable. This finding is consistent with previous research suggesting that cancer often follows a more predictable disease trajectory, which may facilitate more direct discussions regarding prognoses [5, 9, 10]. However, such awareness was not universal even in PCUs, which may reflect the proxy nature of the assessment and the possibility that terminal status is not always explicitly recognized or communicated. Family-reported patient-physician EOL discussions about care settings and code statuses were lower in the hospital groups than in the home groups, regardless of diagnosis. In Japan, where most deaths occur in hospitals, hospitalization may represent the default course of care. In contrast, dying at home typically requires explicit advance discussion and planning. In addition to these structural factors, this pattern may also reflect cultural norms surrounding communication in Japan. Particularly in Japan and in the broader Asian context, the cultural concept of “ishin-denshin” - nonverbal communication - has been described as influencing how sensitive issues are addressed. This mode of communication may partly explain the paucity of patient-physician EOL discussions in hospital settings, as it reflects a culturally embedded respect for maintaining harmony and collective decision-making over individual autonomy [15, 24]. Comparison of quality of death and dying Accurate patient prognostic awareness has been reported to be associated with decreased uncertainty, depression, and anxiety, as well as increased quality of life and improved caregiver mental health [9, 25, 26]. However, our findings challenge the assumption that accurate prognostic awareness directly correlates with improved QOD. This inconsistency is in line with what previous research has shown [12, 13]. Achieving accurate prognostic awareness typically involves discussing disease status and prognosis, followed by specific EOL discussions [27–30]. From a behavioral change perspective, it can be interpreted that in the continuum of serious illness communication, prognostic awareness occurs relatively early, whereas EOL discussions tend to occur later [29]. Therefore, prognostic awareness alone may not be sufficient to achieve long-term outcomes such as improved QOD. This suggests that, while prognostic awareness is important, it may need to be complemented with detailed EOL discussions to enhance QOD fully. In addition, prognostic awareness is multifaceted, but this study only assessed the prognostic aspect of “incurability” and did not consider other aspects, such as imminent death or decline in physical function [29, 31, 32]. Furthermore, reliance on proxy scores complicates interpretation and may not fully capture patients’ subjective experiences [33, 34]. Our notable finding is that differences in family-reported QOD were remarkable in participation in EOL discussions. We found a large difference in the QOD when discussions focused on preferred care settings and code statuses. This finding supports previous research showing that a timely discussion of the code status is associated with less aggressive EOL care and a greater likelihood of receiving palliative care [4, 35]. Although many patients preferred to die at home, similar results were observed for in-hospital deaths. These findings suggest that differences between care settings cannot be attributed solely to cultural factors but may also reflect contextual differences in care processes. Overall, the act of engaging in EOL discussions itself appears to be associated with QOD, regardless of the patient’s place of death. The results also provide interesting insights into the relationship between EOL discussions and QOD in non-cancer conditions, such as heart disease, cerebrovascular disease, pneumonia, and renal failure. In addition, this suggests that discussions on specific topics, rather than just code statuses, may be more beneficial. Discussions about care settings directly related to daily living and patient comfort appeared to be most strongly associated with higher QOD. This observation is consistent with previous studies showing an association between the last place of care and QOD [6, 36]. EOL discussions were associated with greater trust among patients, families, and physicians, which was in turn associated with better overall QOD. Given the nuanced nature of these interactions, especially in high-context cultures such as Japan, this study contributes significantly to our understanding of how EOL discussions may be adapted and culturally attuned to meet the needs of patients and their families across different cultural and disease contexts. Limitations This study has several limitations that warrant consideration. First, the use of bereavement data collected 13 to 25 months after the decedent’s death introduces a potential recall bias. This interval may have affected the accuracy of reported prognostic awareness, as the retrospective perception of family members may differ from the actual patient awareness at the time of death [33, 34]. Given the 54% response rate, non-response bias cannot be excluded. Second, because prognostic awareness is multifaceted, future research should explore which aspects of prognostic awareness are related to specific outcomes and clarify the mechanisms underlying these relationships. Therefore, prospective studies involving such patients are critical. Third, the unique healthcare system and cultural setting of Japan, where this study was conducted, may limit the generalizability of the findings to other contexts. Excluding deaths in long-term care facilities may also limit generalizability to this care setting. Fourth, residual confounding due to unmeasured clinical factors (e.g., comorbidities or access to palliative care) cannot be excluded. Reverse causation related to retrospective proxy reporting may have influenced the observed associations. Given the very large sample size, statistically significant differences may not always reflect clinically meaningful effects. Despite these limitations, the data provide a solid basis for assessing the associations between prognostic awareness and EOL discussions on QOD. Practical implications These findings have practical implications for clinical practice and policy. Given that participation in EOL discussions was more strongly associated with QOD than prognostic awareness alone, clinicians may consider prioritizing timely, structured, and goal-oriented discussions about care preferences and code status across disease categories, including non-cancer conditions. Educational programs may benefit from strengthening communication training beyond oncology settings. At the institutional level, systematic approaches to facilitate EOL discussions, particularly for hospitalized patients with advanced non-cancer illnesses, could help support goal-concordant care. These findings may inform broader palliative care policies aimed at promoting advanced care planning across multiple life-limiting conditions. Conclusion Patient participation in EOL discussions was associated with higher family-reported QOD across disease categories and places of death. These associations were observed not only among patients who died at home but also among those who died in hospital settings. Differences related to discussions about care settings were particularly notable. In this observational study, participation in EOL discussions showed a stronger association with QOD than prognostic awareness alone. Findings from Japan’s high-context cultural setting provide insights into the potential relevance of facilitating structured and culturally sensitive EOL discussions across diverse clinical contexts. Supplementary Information Supplementary Material 1. Acknowledgements The authors would like to express their gratitude to the bereaved family members who participated in the study. Abbreviations End-of-Life Quality of Death and Dying Good Death Inventory Palliative Care Unit Do Not Resuscitate Odds ratios Confidence Interval Authors’ contributions Conceptualization and Methodology: Y.N., M.Mi., T.M., Y.K., A.O.; Investigation: Y.N., E.T., R.Y.; Writing-Original draft preparation: Y.N., M.Mi., M.Mo., T.M.; Writing-Reviewing and Editing: Y.N., M.Mi., M.Mo., T.M., Y.O., Y.K., Y.O., S.K., H.Y., A.O.; Data curation and Visualization: Y.N., M.Mi., Y.O., T.M., Funding acquisition: A.O. Funding This study received support from the mortality follow-back survey project budget of the Ministry of Health, Labour and Welfare, Japan, and a Grant-in-Aid for Scientific Research from the Ministry of Health, Labour and Welfare, Japan (grant number 22EA2001). Data availability The data that support the findings of this study are not publicly available due to regulatory restrictions, but may be made available from the corresponding author on reasonable request and with necessary institutional or ethical approvals. Declarations Ethics approval and consent to participate This study was approved by the National Cancer Center of Japan’s Institutional Review Board (reference number 2017 − 346; approved on June 5, 2018). Consent for participation was provided by checking the consent box and returning the questionnaire. Consent for publication Not applicable. Competing interests The authors declare no competing interests. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. References Associated Data Supplementary Materials Supplementary Material 1. Data Availability Statement The data that support the findings of this study are not publicly available due to regulatory restrictions, but may be made available from the corresponding author on reasonable request and with necessary institutional or ethical approvals.