This multi-method study used the STROBE checklist. Quantitative data were collected using an online questionnaire to understand the level of palliative care attitudes in cancer patients. Qualitative data were collected and coded using open-ended survey questions to explain and validate the root causes of quantitative results. The narrative comments from open-ended questions are typically meant to provide a forum for explanations, meanings and new ideas, the purposes and functions of qualitative and quantitative data in this study on questionnaires are different, yet complementary (22, 23). A convenient sample of cancer patients was selected between March 30 and June 30 of 2024 from the outpatient setting of the Cancer Hospital Chinese Academy of Medical Sciences. The trial population consisted of cancer patients of any type and stage who agreed to participate. The inclusion criteria were: (a) patients aged 18 and older with a confirmed cancer diagnosis, and (b) no noticeable cognitive impairment. Patients who unable to fill out all questionnaires due to physical or emotional stress were excluded. Larger sample sizes in surveys are known to yield more accurate and representative results. According to MacCallum (24), the minimum required sample size of 100, or the sample size to variable ratio should be at least 5. Finally, 541 cancer patients participated and completed the study.

Sociodemographic characteristics divided into two main sections: sociodemographic information, including age, gender, marital status, ethnicity, religion, educational status, occupational status, residence, average family income per month, type of medical insurance, and primary caregivers; the disease-related information, including the duration of illness, cancer stage, level of pain, level of anxiety, level of depression, level of disease knowledge, level of palliative care knowledge, experiences of palliative care-related education or training.

The Palliative Care Attitudes Scale (PCAS-9) was developed by Perry (25) in 2020 and translated by Liu (26) in 2022. It includes 9 items across 3 dimensions: emotional (3 items), cognitive (3 items), and behavioral (3 items). The emotional dimension was reverse scored on a 6-point scale ranging from 1 (not at all stressful) to 6 (extremely, extremely stressful). The cognitive and behavioral dimensions were rated on a 7-point scale ranging from 1 (definitely no) to 7 (definitely yes). The total score ranges from 9 to 60, with higher scores indicating more positive palliative care attitudes. The PCAS-9 in this study has a Cronbach’s alpha of 0.894.

To allow participants to express their attitudes toward palliative care in own words and to help us better understand their perspectives, we included an open-ended question: What is your perspective on palliative care? There was no word limit on responses.

This study was conducted on online, which is called “wenjuanxing,” the largest e-questionnaire platform in China. The online questionnaire was set up so that each IP address could only submit one response to prevent duplicate answers. Our research team contacted the directors of the nursing departments and after explaining the study’s purpose, process and obtaining consent, one of our team members sent the questionnaire URL link to the participants who met the inclusion criteria at the outpatient clinic. The first page of the questionnaire described the study in detail and provided informed consent. Only those who selected “Yes, I agree to join” were allowed to complete the rest of the questionnaire.

Profiles were found using LPA in Mplus 8.3. LPA is widely applied to estimate the number of subpopulations in an example. LPA can employ hypothetical categorical variables to clarify the relationship between exogenous continuous-type indicators, allowing for the evaluation of the relationship between exogenous indicators while maintaining local independence among them. The log likelihood, Akaike Information Criterion (AIC), Bayesian Information Criterion (BIC), adjusted BIC, entropy, Lo–Mendell–Rubin, and bootstrapped likelihood ratio tests were employed to evaluate model fit and identify the optimal number of categories. To determine the best model, and the models from each category’s fitting results were mixed with the indicators.

The quantality data were analyzed using SPSS 26.0. For categorical data, we utilized frequency and composition ratios; for continuous data, we used mean and standard deviation. The χ² test was used to analyze categorical variables between groups. The analysis of variance (ANOVA) was performed for assessing continuous variables across groups. A multivariate logistic regression model was used to investigate socioeconomic variances. p < 0.05 indicates a statistically significant difference.

The qualitative data from open-ended questions were evaluated utilizing content analysis methods to interpret the text data through a coding process, followed by theme identification (27). This approach enables immersion in the data, leading to new insights and category generalization. It is particularly beneficial for analyzing large volumes of textual data (28), which was necessary due to our large sample size and the richness of the open-ended responses. Considering that this is a very time-consuming and task-intensive process, we followed a general inductive approach that allowed us to derive our findings from the most common and dominant themes in the raw data (29, 30) and although the analytical strategy of this approach is not very powerful (30), it’s simple and systematic approach is suitable for the purpose of this study as well as for the quality and quantity of the data. Two researchers independently coded 10 transcripts at a time and met to compute a Kappa statistic and discuss discrepancies in coding. This process was repeated until consensus was reached (Kappa statistic >0.70). The steps of this study were showed in Table 1.

The study was approved by Soochow University (with the codes SUDA20240626H01) and was conducted in accordance with the Declaration of Helsinki. Informed consent was embedded at the beginning of the online questionnaire and participants were prompted to withdraw at any time. Participants completed and submitted the electronic questionnaire indicating that their informed consent had been obtained.

The study included 541 cancer patients, 267 of whom were female (49.4%). The participants’ mean age was 53.37 years (SD = 14.38). More than half of the participants (81.0%) were married, and 87.8% participants at early to mid-stage cancer stage. Table 2 shows the detailed characteristics.

Participants had a total palliative care attitudes score of (27.38 ± 3.59). The scores for each item can be seen in Supplementary Table 1 and Supplementary Figure 1, with Item 5 “Do you think a palliative care consultation would help with feelings of sadness and depression?” rating as the highest, whereas Item 3 “How stressful would you find discussing emotions, like feeling sad, scared, or angry?” had the lowest.

There were statistically significant differences in palliative care attitudes by cancer patients’ ethnicity, religion, education status, occupational status, residence, average family income per month, type of medical insurance, primary caregivers, cancer stage, anxiety, depression, disease knowledge, palliative care knowledge, and palliative care-related education or training experiences (p < 0.05) in single-factor analysis (Supplementary Table 2). Considering the variables that were statistically different in the single-factor analysis as independent variables and the total palliative care attitude score as the dependent variable, the results of the multiple linear regression analysis showed that religion, average family income per month, primary caregiver, and palliative care knowledge were the main influences on cancer patients’ palliative care attitudes (p < 0.05) (Supplementary Table 3).

Using the Palliative Care Attitudes Scale total score as the indicator, the model was fitted to possible profiles 1 through 5 in the present study. According to the p-values for LMR and BLRT, the five-profile model was excluded. The AIC, BIC, and ABIC were decreasing in entropy, indicating that the profile model performs better as the number of profiles increases. The entropy value is higher in four-profile compared to three-profile and the probability of each category is greater than 5% for the four-profile model, the four-profile model was found to be the best suited model with high classification certainty, taking into account the previously mentioned data (Table 3 and Figure 1).

Class 4 presented the highest palliative care cognitive and behavioral scores but lowest emotional scores (named the ‘low-emotional, high-cognitive & behavioral group’). Class 3 showed low level of cognitive and behavioral scores but highest emotional scores (named the ‘high-emotional, low cognitive & behavioral group’). Class 2 had the lowest emotional scores (named the ‘low-emotional group’). Class 1 had the overall more even and not too high scores (named the ‘balanced negative group’).

Multivariate logistic regression result was showed in Table 4. The profiles of palliative care attitudes were used as the dependent variable, with the “low emotional, high-cognitive & behavioral group” as the reference. Sociodemographic information and disease-related information served as independent factors in the analysis. The findings revealed a statistically significant effect of education status, occupational status, primary caregivers, type of insurance, cancer stage, anxiety, and level of palliative care knowledge on the different profiles of palliative care attitudes (p < 0.05).

Cancer patients’ perspectives on palliative care were divided into three themes: “theme 1, expectations for the promotion of palliative care services (21.7%),” “theme 2, skepticism and refusal toward palliative care development (41.2%),” “theme 3, indifference to palliative care due to limited knowledge (37.1%).” Figure 2 shows the frequency of occurrence of each theme, the most common perspective is theme 2. We discuss each theme according to its order of frequency, with example statements from responses.

In qualitative analysis, the most common theme identified was “skepticism and refusal toward palliative care development.” Fifty-eight (10.7%) patients explicitly rejected palliative care. “I’ve been to many hospitals and tried numerous treatment plans; I will not believe in any new methods anymore,” “There is no palliative care available in township hospitals, which is inconvenient,” “The cost must be high, and I do not want to waste money anymore,” and “I’ve never heard of palliative care before, are you trying to experiment with my life?.” Additionally, 151 (27.9%) patients doubt palliative care, asking “What is palliative care, and can it really prolong my life?” “Is palliative care treatment or care?” “Can palliative care relieve my pain?” “As long as it can reduce my depression, I’m willing to try,” “Where can I consult about palliative care?” “Can palliative care consultation be done online?” and “Is palliative care euthanasia? I’m scared when I hear that word.”

The second frequently encountered theme is “indifference to palliative care due to limited knowledge.” In the final phase of the quantitative study, an open-ended question was posed, allowing participants to provide any text-based responses, including thoughts, suggestions, and queries. However, 140 (25.9%) participants answered only “none” and 62 (11.5%) participants with “I have no idea because I have never known palliative care before.” The remaining minority responded with “First time hearing about palliative care.”

Although Theme 3 implies a positive attitude toward palliative care, it has the lowest frequency. Of the 69 (12.7%) participants, a desire to engage in palliative care was expressed. Responses such as “I wish to experience palliative care” “I am interested and would like to seek palliative care consultation” and “This is my first encounter with palliative care, and I will arrange a consultation promptly” were common. 36 (6.7%) participants mentioned their desire to further understand palliative care and intention to pay attention to it. 15 (2.8%) participants expressed belief in palliative care, stating sentiments like “I hope palliative care becomes more widespread, so I can benefit from it even in rural hospitals” and “I heard palliative care before, it is a comprehensive treatment, and I hope it continues to improve.”

The findings of this study provide significant insights into cancer patients’ poor attitudes toward palliative care in China, emphasizing the need for further exploration and intervention. Future research could investigate additional psychosocial variables, such as cultural beliefs, religious affiliations, and social support systems, which may further elucidate the complexity of palliative care attitudes around worldwide. Additionally, future research should develop and evaluate targeted educational interventions, the focus of which can be tailored to the patient’s education level, occupational status, level of palliative knowledge, etc., to accurately raise public awareness. By doing so, researchers can develop evidence-based strategies to enhance public acceptance and ultimately improve the quality of end-of-life care.

There are several limitations. Firstly, there is a limited number of young and middle-aged cancer patients among the participants. Although the sample size is a strength, the uneven distribution of age groups may weaken the representativeness of the sample. Secondly, the textual responses were collected from qualitative questions in an online survey. Despite many responses being comprehensive, they still lack the depth that qualitative study can produce. We suggest further conducting rigorous qualitative research to deepen the study results. Lastly, we performed frequency calculations on the themes identified in the qualitative research results. Due to some responses being more elaborate, they were coded multiple times, which may lead to some responses being overrepresented in the themes.