To the best of our knowledge, this is the first study to concurrently use DCE and BWS to reveal Chinese residents’ preferences for COVID-19 vaccination and to compare the differences in the results of these two methods. Our findings indicate that Chinese residents prefer COVID-19 vaccines that are domestically branded, have higher effectiveness, longer duration of protection, lower risk of severe adverse events, and are administered orally. In addition, our results indicate a similar pattern in the DCE and BWS methods, with the respondents having a strong preference for 90% vaccine effectiveness. However, the methods diverged in other preferences. When comparing the results with existing research, the discrepancy suggests that DCE is more suitable for capturing preferences in our context.

Understanding Chinese residents’ preferences for COVID-19 vaccines could significantly enhance public health strategies by informing targeted vaccine communication and distribution plans, ultimately increasing vaccination uptake and contributing to improved public health outcomes. In terms of brand, our finding shows the domestic vaccine is preferred, which is consistent with a previous study [39]. It is believed that the government’s promotion and advocacy of domestic vaccines during the COVID-19 outbreak has increased public trust in them [40, 41]. Moreover, Chinese people are willing to pay 200/400 CNY for the vaccines, which shows their understanding of the importance of vaccination and their purchasing power to some extent. It is important to note that the hypothetical scenarios differ from the fact that the COVID-19 vaccine in China is provided for free by the government and health insurance, which may introduce a potential bias in the study.

The results of the comparison between DCE and BWS indicate that the overall consistency is poor, supporting the hypothesis that these methods may not yield equivalent findings. In other studies of stated preferences, the findings from Whitty et al. [42], Soekhai et al. [43], and Armeni et al. [44] also indicate a relatively low level of consistency between the two methods. Specifically, in terms of attribute ranking of the whole population, the attribute ‘effectiveness’ ranks second in the DCE and last in BWS. This is noteworthy because previous studies of COVID-19 vaccines [10, 15, 16, 18, 23, 45,46,47], and even studies of other vaccines like those for influenza [48] and HPV [49], have identified effectiveness as one of the top three most important attributes. Our study found that even with various methods, it is still impossible to achieve complete orthogonality when the number of attributes and levels is large, resulting in a trade-off phenomenon. In other words, in the selection scenario, when the valued attribute is present, it may be traded off with other attribute levels, where a less attractive level of the valued attribute will be chosen to be the least important, ultimately affecting the overall ranking of the relative importance of the valued attribute. For instance, in this study, even though effectiveness is considered the most important attribute, when a 60% effectiveness occurs more frequently, this level is highly likely to be selected as the least important, thus affecting the overall relative importance score ranking of the attribute ‘efficiency’. In other words, this is related to the fact that BWS makes it easier to identify and express extreme preferences, which also leads to the possibility that it may not adequately capture participants’ more complex trade-offs at the attribute level, resulting in bias between its preference estimates and DCE [50]. This discrepancy suggests that DCE may offer a more practical and suitable method for capturing preference for COVID-19 vaccine selection. Besides, it is suggested that BWS is suitable for simpler situations where properties and levels are limited and can be completely orthogonal, while DCE is a better choice in more complex product selections. Our research results are in line with studies by Soekhai et al. [43], Flynn et al. [51]. and Himmler et al. [52], in which the authors also reported that DCE may result in less cognitive burden than BWS-2, making it more appropriate for decision-making contexts that require the evaluation of multiple attributes and levels.

Another possible reason for the lack of concordance between the two methods is the differences in preference construct. DCE and profile-case BWS are grounded in distinct theoretical frameworks and psychological models. This may lead to variations in how participants express their preferences, resulting in different characteristics and patterns within the same research context. Consequently, there may be fundamental discrepancies between the two methods regarding the concordance of preference information acquisition, as noted in the literature by Whitty and Oliveira Gonçalves [53]. However, it is not conclusively established, and further research is warranted to explore these findings in greater depth. In addition, the results of the BWS analysis revealed that both the vaccine hesitancy group and the vaccine recipient group assigned the same relative importance score ranking to various attributes. We believe this might be due to the limitations of the BWS method, which may not effectively capture and distinguish preference differences between these groups.

Further, we found there is preference heterogeneity for COVID-19 vaccination among different groups. Both the vaccine-hesitant group and the high-risk group prefer oral vaccination, which indicates that the availability and convenience of vaccines are also crucial selection conditions [54]. This may be related to the misconception that the traditional intramuscular injection has more adverse reactions after vaccination and might seriously affect daily life [55]. Another latent reason might be the fear of needles, which can produce nervousness, anxiety, and other uneasy emotions [56]. Moreover, traditional intramuscular injection vaccination needs to be carried out by medical professionals, and a lack of convenient access, unfamiliarity, long waiting times, and other obstacles may lead to the rejection of intramuscular injection of COVID-19 vaccine [55]. Additionally, oral vaccination is an ideal and easy-to-accept delivery route, which is highly accepted by residents, and the oral method is convenient, allows self-management, and is suitable for people of all ages [56, 57]. The preference for oral vaccines has significant implications for improving targeted vaccination strategies and reducing vaccine hesitancy.

In addition, we found that for the vaccine recipient groups, emphasizing the ‘added’ properties of vaccines has an important role. For example, the promotion of vaccination methods and vaccine brands should be highlighted because this group usually already has a high level of identification and trust in the basic attributes of vaccines (such as effectiveness, duration of protection, etc.). For the vaccine-hesitant group, however, the picture is different. In addition to promoting vaccination methods to reduce psychological barriers, it is necessary to carry out targeted publicity on the core attributes of vaccines, especially emphasizing the effectiveness of vaccines and the low risk of adverse reactions in order to enhance the acceptance of vaccines in this group.

This study has several notable advantages. Firstly, the attributes and levels of the DCE and BWS were established based on a thorough literature review and two rounds of Delphi consultations, providing a solid foundation for the experiment. Secondly, while both DCE and BWS are widely used methods to estimate preferences, they are grounded in different theoretical frameworks and involve distinct choice architectures. This study explored the potential discrepancies of DCE and BWS measurements in the field of preference for COVID-19 vaccination, providing evidence for improving the accuracy of preference assessments. Thirdly, vaccination methods like oral, injection, and aerosol inhalation, which were relatively rare in previous studies, have been included. In addition, our subgroup analysis provides additional insights for preferences of different groups.

This study has some limitations that need to be acknowledged. Firstly, it only focused on the first six attributes that experts considered most important in using the DCE and BWS methods. It was impossible to include a more comprehensive combination of attributes and levels, which may have led to other factors that could affect vaccination preference being overlooked. Secondly, the online survey was conducted through convenience sampling, which may not have accurately represented the characteristics of the sample population. Certain demographic groups, such as the elderly, who may not be as inclined to fill in an online questionnaire, might have been excluded, possibly resulting in sample selection bias that could influence our analysis. Thirdly, to maintain consistency with BWS settings and ensure comparability of results, we did not include an opt-out option when designing the DCE. While this approach aligns with our objective of evaluating participants’ preferences for vaccine attributes, it may have constrained our ability to fully capture the preferences of individuals who would have chosen not to vaccinate. However, subgroup analysis in this study provided additional insights by revealing significant differences in preferences between the vaccination and vaccine hesitancy groups. Through this, we partially address the potential limitation by identifying the preferences of the vaccination hesitancy group, providing valuable information that might otherwise have been missed. Nonetheless, we acknowledge that not including an opt-out option might have influenced some participants’ decision-making process, and this should be considered when interpreting the results. Moreover, differences in how the vaccination and vaccine hesitancy groups understood the selection tasks and expressed their preferences may have introduced bias, potentially deviating from an ideal comparative framework. Future research could address these limitations by incorporating opt-out options into the experimental design. This adjustment would better reflect real-world scenarios, particularly in contexts where vaccination is not mandatory or strongly encouraged, thereby enhancing the external validity of the findings. Finally, this study is subject to hypothetical bias, as respondents are required to make choices between hypothetical service options rather than objective reality. Despite these limitations, this study provides a novel perspective to understanding COVID-19 vaccine preference comparing the DCE and BWS, offering valuable insights that have not been previously explored.