A prospective cohort study was adopted between January 2015 and August 2023 to collect data from terminal cancer patients in Wuhan, China, through face-to-face interviews and telephone-based surveys with questionnaires.

Terminal cancer patient-family caregiver dyads were recruited from four tertiary hospitals in Wuhan, China, via a convenience sampling method. Our research team members contacted the nursing administrators of four hospitals and explained the purpose and procedures of the study. The eligibility criteria for the participants included patients diagnosed with terminal cancer, unresponsive to curative therapy, competent in cognition, whose estimated life expectancy was less than 6 months, and during hospitalization. Family members identified as the primary caregivers to be involved in patients’ EoL care, able to communicate coherently, aged 18 and older, were also recruited. The exclusion criteria specified patients or caregivers who were not Chinese.

Given that most Chinese work from Monday to Friday and rest on weekends in a week, it will be less busy on weekends, so the data were collected at 9–12 a.m. and 2–5 p.m. on weekends from 1 January 2015 to 31 August 2023 to allow more flexibility and respondence. After the eligibility screening in the oncology department, terminal cancer patients and family caregivers who signed informed consent or gave oral consent were face-to-face interviewed separately in an intimate room by a trained research nurse first to collect their socio-demographic information and patients’ preferred PoD. The follow-up telephone interview took place in the last month of the patients’ lives, estimated by the physician in charge to learn the patients’ preference for PoD again. At last, within 1 month after the patients’ death, families were interviewed about patients’ reception of hospice care and actual PoD by phone.

Ethical approval of the research protocol was granted by the institutional review board of Huazhong University of Science and Technology [registration number: [2019]IEC(S198)] as well as all the institutional review boards of the participating hospitals. All participants were informed of their voluntary participation in the study and their right to refuse to participate and withdraw from the study at any time, and all information was kept confidential.

Data analysis was calculated with SPSS 24.0 (SPSS Inc., Chicago, IL) (IBM Corp, 2012). Descriptive statistics were used for participants’ characteristics. The κ statistic was used to assess the agreement between patients’ preferred and actual PoD, with values ≤0 as indicating no agreement, 0.01–0.20 as none to slight, 0.21–0.40 as fair, 0.41–0.60 as moderate, 0.61–0.80 as substantial, and 0.81–1.00 as almost perfect agreement (McHugh, 2012). Binary logistic regression models were applied to identify how significant variables were associated with the congruence. Differences at a p-value of <0.05 were demonstrated as statistically significant. An OR with a 95% confidence interval (CI) was calculated for each outcome variable.

The baseline characteristics of 845 dyads are presented in Table 1. Approximately half (52.3%) of the patients were male, with a median age of 54.1 years old. The most common type of cancer was lung cancer (27.5%), followed by breast cancer (17.7%), head and neck cancer (14.6%), intestinal cancer (13.0%), and gastric carcinoma (8.2%); and cancer had lasted for approximately an average of 422 days from diagnosis to study enrollment. Family caregivers were mostly the patients’ spouses (59.2%), followed by sons or daughters (27.9%). Approximately half of them were female (55.0%) with an average age of 47.2 years old.

Most dying patients had pain (69.0%) and did not receive hospice care (89.1%). A total of 27.6% of the family caregivers rated the quality of care received by patients in the last 48 h as good, while 55.4% of them thought patients’ quality of death was good (Table 2).

For patients’ preferred PoD, the home was selected by most patients (83.7%) in the baseline interview but decreased (68.1%) in the follow-up interview in the last month of their life. The percentage of the preference for dying in hospitals and hospice facilities increased from 9.3 to 24.4% and from 0.7 to 6.1%, respectively, in the two interviews. Additionally, the agreement in patients’ preferred PoD in patient-caregiver dyads was both fair in the baseline interview (κ = 0.211, p = 0.000) and in the follow-up interview (κ = 0.267, p = 0.000) (Table 3).

For patients’ actual PoD, most patients died in hospitals (62.7%), followed by home (36.1%) and on the way to hospitals (1.2%). The congruence between patients’ preferred and actual PoD was 63.0, 36.6%, and 0 for hospital death, home death, and hospice facilities/nursing homes/relative’s home death separately (Figure 2). In summary, the agreement between patients’ preferred and actual PoD was fair (κ = 0.221, p = 0.000) (Table 4).

The congruence between patients’ preferred and actual PoD was more likely to be achieved when patients’ preferred PoD and family caregivers’ belief of patients’ preferred PoD was congruent (OR = 6.464, p = 0.001), or caregivers had a medical-related occupation (OR = 4.577, p = 0.003); if hospitalizations occurred at least twice in the last month of patients’ lives (OR = 0.422, p = 0.000), or the quality of care received by the patients in the last 48 h was good rated by the family caregivers (OR = 0.373, p = 0.011), patients were less likely to die in their preferred places (Table 5).

Consistent with previous findings (Agar et al., 2008; Hinton, 1994), our study found that terminal cancer patients’ preferences for home death decreased, while the preference for dying in hospitals or hospice facilities increased as death approached. When dying is approaching, preparation for symptom control related to dying is required (National Cancer Center and National Hospice Center, 2020), while families are less confident in offering quality home care (Gao et al., 2013). In addition, this study also collected data during the COVID-19 pandemic. Considering the limited professional support outside hospitals during the COVID-19 wave, no wonder many people wanted to go to hospitals for symptom control (Lucijanic et al., 2023). Hospice care aims at relieving patients’ bad sufferings and helping them acquire a relatively comfortable dying process with dignity (Kumar et al., 2017), but such services are mostly provided in hospitals in China as professional hospice facilities are in great shortage in most areas of China, including Wuhan (Lu et al., 2018). Patients and families believe hospitals are more reliable to deal with discomfort symptoms than other settings (Li et al., 2020). Suggesting hospice services need to be expanded, not only in hospitals and hospice facilities but also provide hospice at home or nursing homes under potential legislation for dying patients to make flexible choices of preferred PoD.

The agreement between patients’ preferred and actual PoD was fair. A total of 63.0% of those patients preferred hospital death and 36.6% of those preferred home death ultimately died in their preferred places, but the percentage was 0 for those who preferred dying in hospice facilities/nursing homes. Impacted by the Chinese principle of filial piety, compared with those patients who preferred to die in other places, patients’ wishes for hospital death were more likely to be met, as their families were more likely to make all efforts to prolong the patients’ life (Ling et al., 2020). For the home death preference, patients and families who acknowledged the Chinese proverb “falling leaves return to their roots” were more likely to return home when death was approaching (Li et al., 2020). The preference for hospice death or nursing home death was difficult to follow due to the unawareness and inaccessibility to hospice services (Wu et al., 2016; Yin et al., 2017). Additionally, hospice care fees are only partially covered by medical insurance in China (China Health Insurance, 2020); families need to pay for hospice care out of their pocket, while 57.4% of patients’ family financial status was not adequate in our study, which may be another reason why the percentage was 0 for those who preferred dying in hospice facilities/nursing homes. Additionally, the COVID-19 pandemic might also influence the changes in PoD. Shibata et al. found home death appeared in a strong uptrend while hospital death exhibited downward trends after 2019 (Shibata et al., 2024), as many were without any option stuck at home despite their wishes to be in hospital or hospice, leading to a decreased congruence between preferred and actual PoD. The fair agreement between the preferred and actual PoD suggests a need to establish and utilize community-or home-based hospice services and increase the reimbursement of hospice care under potentially considering legislation to support advance directives.

If patients’ preferred PoD and family caregivers’ belief of the patient’s preferred PoD was the same, patients were more likely to die in their preferred places. Previous literature has indicated that the inconsistent gap between patients and their surrogates regarding EoL care preferences can be decreased through ACP (Ke et al., 2021); in addition, ACP indeed helps patients receive care concordant with their preferences (Detering et al., 2010). Given the fact that ACP remains unfamiliar to the public and has not been regulated by legislation in China (Zhang et al., 2021), it is urgent to establish relevant policies or guidelines to promote the awareness and development of ACP in China, to increase the congruence between patients and caregivers about patients’ PoD preferences, and thus to promote the compliance of patients’ EoL care preferences.

If the family caregivers engaged in the medical-related work, patients were more likely to die in their preferred places. Family-centered decision-making process is admired in Asian culture (Haley et al., 2002). Family caregivers with medically related occupations may tend to occupy the leading position in the decision-making process and are more willing to initiate death topics with patients (Itzhaki et al., 2016), which may make them know more about the patient’s preferred PoD. Furthermore, they might be more aware that satisfying patients’ preferences was an essential indicator of the quality of good death (Peláez-Cantero et al., 2023); thus, they were more prone to respect and follow patients’ preferences. This highlights that detailed ACP or EoL care discussions should be started earlier to promote the families’ awareness of patients’ EoL preferences.

When patients were hospitalized at least twice in the last month of life, they were less likely to die in their preferred places. The more hospitalizations suggest patients may suffer more emergent symptoms that need to be addressed. Under the healthcare background in China, hospitals, especially tertiary hospitals, are the priority for patients and families to seek professional help, as most community hospitals or nursing homes can only offer basic medical care to various patients (Gu et al., 2015). Other reasons may include the less utilization of hospice facilities (Lu et al., 2018), insufficient medical resources, and weaker professional competence of nursing home staff (Gu et al., 2015), leading the patients to go to hospitals for treatment. It is important to reallocate and balance the medical services among hospitals, hospice facilities, and nursing homes, to make patients receive timely treatment in their preferred PoD.

If family caregivers considered patients had received high-quality care in the last 48 h, patients were less likely to die in their preferred places compared to those who acquired low-quality care or did not receive care. Caregivers may wish the high-quality care to last till the last moment of patients’ lives, resulting in increasing numbers of hospital deaths, whatever the patients’ preferred PoD was. For the patients who received high-quality care in hospitals in the last 48 h, they might change their preference from other places to hospital death, but this actual “last” preference was unknown to us as almost all the patients were unconscious or could not express normally at this EoL stage, suggesting to note and update patients’ EoL care preferences more dynamically. It is urgent to set up more hospice facilities in China to provide high-quality care not only in hospitals but also to improve the quality of care in hospice facilities, nursing homes, and at home, thus meeting patients’ EoL care preferences with flexibility.

Our study provides insight into the congruence of patients’ preferred and actual PoD and the factors of it, as well as the agreement in patients’ preferred PoD in patient-caregiver dyads among terminal cancer patients in China for the first time with a relatively large sample size, which will help to improve compliance of patients’ EoL care preferences. There are several limitations. First, a convenience sampling method was used to recruit participants from four tertiary hospitals in Wuhan, China, possibly limiting the findings to be generalized to patients in primary or secondary hospitals, and in communities or medical care facilities. Second, patients’ preferred PoD interviewed in their last month of life may not be totally the same as the preference in the last few hours. Future research with a mixed method in different facilities could be carried out to provide more information on the congruence of patients’ preferred and actual PoD.