Public preference for COVID‐19 vaccines in China: A discrete choice experiment

Abstract Background As the coronavirus disease 2019 (COVID‐19) pandemic is sweeping across the globe, there is an urgent need to develop effective vaccines as the most powerful strategy to end the pandemic. This study aimed to examine how factors related to vaccine characteristics, their social normative influence and convenience of vaccination can affect the public's preference for the uptake of the COVID‐19 vaccine in China. Methods An online discrete choice experiment (DCE) survey was administered to a sample of China's general population. Participants were asked to make a series of hypothetical choices and estimate their preference for different attributes of the vaccine. A mixed logit regression model was used to analyse the DCE data. Willingness to pay for each attribute was also calculated. Results Data of 1236 participants who provided valid responses were included in the analysis. There was strong public preference for high effectiveness of the vaccine, followed by long protective duration, very few adverse events and being manufactured overseas. Price was the least important attribute affecting the public preference in selecting the COVID‐19 vaccine. Conclusions The strong public preferences detected in this study should be considered when developing COVID‐19 vaccination programme in China. The results provide useful information for policymakers to identify the individual and social values for a good vaccination strategy. Patient or Public Contribution The design of the experimental choices was fully based on interviews and focus group discussions participated by 26 Chinese people with diverse socio‐economic backgrounds. Without their participation, the study would not be possible.


| INTRODUC TI ON
As of 24 August 2020, the novel severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) virus has infected more than 23 million people in 216 countries and regions, with a case fatality ratio (CFR) of approximately 3.4%. 1 Currently, there is no effective treatment for this disease, and relaxation of effective non-pharmaceutical interventions often leads to resurgence of community outbreaks. [2][3][4][5] Thus, a vaccine seems to be the only solution to this problem. As vaccines are regarded as the most cost-effective way of controlling infectious diseases, there are attempts to develop a coronavirus disease 2019  vaccine rapidly to catch up with the rate of the pandemic's spread. 6 On 20 July 2020, the so-called Oxford vaccine (ChAdOx1 nCoV-19) was announced as a front runner among 24 candidate vaccines in clinical evaluations worldwide. The reason is that it was proven in a stage 1/2, single-blind, five-site, randomized clinical trial that the vaccine could spike up antibodies and create a safe immune response in the body. 7 However, the path to introducing a new vaccine to the market can be politically and economically complicated. The  vaccine is no exception. Although insights and opinions of different stakeholders-such as policymakers and medical professionalsmight affect the vaccine's uptake to some extent, 8 the most essential factor for any vaccination programme's successful adoption is the public's acceptance. Factors such as individual characteristics (eg high-risk occupations and socio-economic vulnerability) and disease-specific characteristics (eg morbidity) play an important role in the individual's decision-making process to select vaccination programmes. 9,10 A reasonable strategy should consider both the provider's affordability and consumer's preference. However, currently, studies investigating these factors and their effect on the public's preference in selecting the COVID-19 vaccine are limited and fragmented. Obtaining such information is important for policymakers to understand the individual and social values to optimize strategies and design potential vaccination campaigns to address COVID-19 as well as for pharmaceutical companies to estimate the expected benefit when managing the vaccine's development. 11 Moreover, the acceptance rate for a possible vaccine also reflects the public's willingness to be vaccinated. Thus, this study aimed to examine how the relative importance of factors related to vaccine characteristics, the social normative influence and convenience of vaccination affects the public's preference for the uptake of the COVID-19 vaccine in China.
China was selected as the research location for two reasons.
First, China is one of the leading candidates in the global vaccine development contest, as three of its vaccines were reported to have already completed the phase 2 clinical trials. At the end of June 2020, China's state-run CanSino Biologics announced that their vaccine candidate demonstrated a 'good safety profile' with high levels of immune response in patients, and it is highly probable to be authorized for urgent use, including for front-line medical professionals, at the end of this year. 12 Thus, a broader commercial use of the vaccine may not be too far off. Second, China's Wuhan City is regarded as the epicentre of the COVID-19 pandemic. Moreover, China is one of the few countries recovering from the pandemic via careful manoeuvering to return to normal. Nevertheless, the pandemic's impact at the physical, psychological, social and economic levels is extensive and long-lasting. Hence, this study provides empirical evidence to identify the optimal COVID-19 vaccination programme for promoting the vaccine's uptake among the general Chinese population and indicates key attributes for consideration when other countries start to develop their own COVID-19 vaccination programmes.

| ME THODS
To explore public preferences for the COVID-19 vaccination programme, we used a discrete choice experiment (DCE) task administered online. 13 For each choice task, there were two options of hypothetical vaccination programme alternatives. To ensure all respondents make a choice and to detect their preference, no opt-out alternative was pro- vided. An example choice set is provided in Figure 1. The major benefit of using the stated preference method is that it allows us to understand and capture the public's preference for vaccination programmes that do not currently exist but could in future be available.

| Selection of attributes and levels
The criteria defined by Norman et al 14 were followed to develop the attributes and levels of our DCE questionnaire. According to these criteria, first, all levels and their combinations should be reasonable.
Second, all levels and their combinations should be familiar to respondents in their current practice. Third, heterogeneity of the levels should be fully considered in the design to ensure the respondents can make some trade-offs between them.
The attributes and levels were selected using a standard iterative process adopted by previous studies that used DCE. 13,15,16 First, the research team conducted a comprehensive literature review with articles extracted from the Cochrane Library, Web of Science, MEDLINE andEMBASE (1950-2019), including academic articles using a variety of research methods such as quantitative and/or qualitative study designs, systematic reviews and randomized clinical trials, and the other government reports and policy briefings from Google, to explore important factors that affect the public's willingness and attitude when making decisions on vaccination uptake. The search target was not limited to the COVID-19 and other pneumonia vaccines, but extended to other fields such as the influenza vaccine. Two researchers

K E Y W O R D S
Chinese public, COVID-19 pandemic, discrete choice experiment, vaccine, willingness to pay independently completed the literature review. All the team members discussed the findings, and four attributes-effectiveness, 17-20 protective duration, 17,19-21 adverse events 22-24 and frequency of injections 25,26 -were confirmed that potentially important for developing our DCE questionnaire. Then, on the basis of the findings from the literature review, six one-to-one interviews (three males and three females, aged between 28 and 62 years) and three focus group interviews with six to eight participants in each group (20 participants in total) as a sample of the Chinese general population were conducted to investigate their views and perspectives about the attributes of an 'ideal' COVID-19 vaccine and its effectiveness at different levels.
Several new insights were derived from the qualitative interviews.
A new attribute-place of origin-that was considered to indirectly reflect the quality of the vaccine was added based on the interview results. The expression and explanations of the attributes and levels were modified according to the interviewees' suggestions and comments. Third, a team of experts from relevant areas (clinical medicine, methodology, infectious disease and policy, five persons in total) was invited to discuss the findings from the previous steps. Two rounds of discussion were organized, experts and research team worked together to discuss how to modify and refine the attributes and levels to meet our criteria. Considering the majority of the adult vaccines are not free of charge in China, a cost parameter, which reflected the price of COVID-19 vaccination programme, was included in the DCE questionnaire. An optimal mode of presenting choice sets to the respondents was also determined by experts to ensure the maximization of the face validity-the extent of a measure to capture what it is intended to measure-of the choice task. 27 Thus, finally, based on the literature review, general public interviews and expert discussions, six attributes with two to six levels for each were developed. The final set of attributes and levels is presented in Table 1.

| Experiment and questionnaire design
A full-factorial design using all the attributes and levels results in Before the formal study, a pilot DCE survey was conducted. A convenience sample of 10 members of the general public was invited to participate in the online survey. First, they were asked to complete the questionnaire independently through the same online platform as in the formal survey. Second, an interview was conducted immediately by the first two authors to understand their comments and suggestions on the survey and the approach.

| Sample and survey administration
The survey was managed online via Wenjuanxing (WJX, https:// www.wjx.cn/), the biggest online survey company in China, between June and July 2020. The questionnaire was developed by the research team using WJX's survey design software built on its online survey platform. Participants were recruited by the same survey company via its members on the online panel.

| Data analysis
Descriptive statistics were used to present the participants' demographics, SES, and physical and mental health status. The random utility theory provides the theoretical foundation for analysing the DCE data. The public's utility (U) associated with a particular vaccination programme had two components: the deterministic component (V) and the stochastic component (ε).
The model of utility for an individual n associated with vaccination programme i can be estimated as The DCE data were binary, where '1' indicates that the alternative plan was chosen and '0' means that the other alternative plan was chosen. All attributes were dummy-coded, and the coefficients of each level were estimated in the model and summarized to reflect the overall utility for each profile. The mixed logit regression (MXL) model was used to analyse the DCE data, as it estimates a distribution around each mean preference parameter to avoid potential bias of the estimated mean preference weights caused by unobserved heterogeneity. 33 The attribute of 'price' was specified as a continuous variable to facilitate the calculation of willingness to pay (WTP), which is the monetary value that people place on different attributes of the vaccination programme.
We calculated the utility value and relative predicted probabilities for all profiles of the experimental design, which allowed us to compare profiles that are more likely to be chosen by respondents with profiles that are less likely to be chosen. 34 This allowed us to convey the DCE results as easily understood information for the general public and policymakers. Subgroup analysis was also conducted to estimate the public's preference heterogeneity regarding vaccination programmes in terms of the respondents' gender (men/ women), family registry (urban/rural), parenting (yes/no) and personal vaccinated experience (yes/no). All statistical analyses were conducted using R (R Foundation, Austria) and STATA. The P-value was set at ≤.05.  Table 3 shows that the order and signs of all the attributes were as expected, and the coefficient of the attributes, except for the 'number of injections = 2', was statistically significant. The results demonstrated that the most important attribute was effectiveness.

| Results of the main effect model
The coefficient of '90% effectiveness' was 3.138 (P < .001), followed by that of '70% effectiveness' (b = 1.416, P < .001). Although the COVID-19 vaccine's price had a negative and significant effect on the respondents, it did not appear to be as important as the other attributes (b = −0.002, P < .001). Respondents' preference for choosing a COVID-19 vaccination programme increased with a longer protected duration but decreased with more adverse events and higher frequency of injections. In addition, we found that the place of manufacturing of the COVID-19 vaccine affected the respondents' preference-imported vaccine generated a higher utility score (b = 0.178, P < .001).
Results of the WTP estimation supported the comparisons of the respondents' preferences from the monetary perspective. The results demonstrated that respondents prefer to pay more for effectiveness and longer protective duration than for the other attributes.
On average, respondents were willing to pay around 1948 CNY and 446 CNY to take vaccines with 90% effectiveness and a protective duration of 18 months compared with 50% effectiveness and a protective duration of six months, respectively. In terms of the frequency of injections, respondents were willing to pay only 98 CNY to take one shot rather than take three shots. Table 4 and Figure 3 present results of the selective subgroup analysis. The COVID-19 vaccine with higher effectiveness was more likely to lead to a higher utility value for respondents who were women, lived in a rural area, parenting children and had vaccinated experience. The utility values and probability of selection for all design profiles are presented in the Appendix (Table A2). 19 vaccine would not be a one-off effort, but a long-term process.

| D ISCUSS I ON
The cost of providing a free COVID-19 vaccine to the public season by season would then be a significant financial burden and an impossible mission for some developing countries. Therefore, our WTP estimations provide useful information for policymakers to develop a reasonable pricing strategy to commoditize the COVID-19 vaccine in the market. In addition, we should not neglect the effect of 'free-riding' behaviours, which were reported by previous studies about vaccination decisions. 16,43,44 Price is likely to only have a slight influence on individuals' vaccination preference, not because they do not care about the cost of vaccination but because they would not get vaccinated and hope to be covered by herd immunity. Herd immunity is developed when other people take the vaccine and create a sufficiently high coverage to protect everyone.
Further, the price of vaccination in our study was limited to five levels, and the public's decisions on the choices might be affected by this predefined price range. However, at the time of conducting the study, no COVID-19 vaccine was available in the market. The The subgroup analysis further demonstrated that the female respondents were more likely to select a COVID-19 vaccine with higher effectiveness, longer protective duration, fewer adverse events and fewer injections than the male respondents. However, the females' preference for vaccination seemed to be more sensitive to increased price. Although previous studies indicated that females are more likely to take up other vaccines than males, 45,46 none discussed the effect of price on the decision making between males and females. Some possible explanations for this might be differences in the SES and health status, and provider bias.    Third, compared with CV, DCE provides better opportunities for researchers to identify people's trade-offs between different attributes of a product. 56 However, it is worth noting that DCE usually generates a higher cognitive burden than CV, especially when the design of a DCE is complicated or the sample size is a relatively small one.

| LIMITATI ON S OF THE S TUDY
Several limitations of our research must be addressed. First, our data were collected from an online survey, which means that people who did not have access to the Internet were excluded from the survey, which is likely to lead to a selection bias. Second, compared with the Chinese general population, our sample is much younger, better educated and has a higher income. Nearly 80% of them reported having an average monthly income greater than the national median. Methodologically, an inherent characteristic of DCE is that respondents have to make a choice between two hypothetical profiles. However, in the real world, they might be presented with more options. Hence, the generalizability of our findings is limited. Third, the low utility of adverse events in our study might be resulted from setting up the range of adverse events at relatively milder levels in the first place. Furthermore, although explanations on the attributes and levels of the profiles were provided in the survey, some participants might not read them carefully or even misunderstood the profiles. Therefore, the validity of our findings is not without concern. Finally, a more heterogeneous approach is needed in future studies by including different stated preference methods such as CV, or statistical techniques such as hierarchical Bayes.

| CON CLUS IONS
This study found that 80% of the Chinese public who participated in the survey preferred to receive the COVID-19 vaccination when it is available. More than 40% of them indicated that the elderly should be prioritized for the vaccination programme. When the participants were facing trade-offs between two COVID-19 vaccination programmes, effectiveness was regarded as the most important attribute, followed by long protective duration, very few adverse events and being manufactured overseas. Interestingly, price was the least important attribute affecting the public preference in selecting the COVID-19 vaccine.
However, such findings need to be interpreted with caution.
The distribution of income levels among our sample was skewed towards the higher end of national average. The public with lower incomes who will be more sensitive to prices was in fact unrepre-

ACK N OWLED G EM ENTS
The authors would like to thank participants for interviews and focus groups for sharing their perceptions of and experience with vaccines.

CO N FLI C T O F I NTE R E S T
All the authors declare no conflict of interest.

DATA AVA I L A B I L I T Y S TAT E M E N T
The data that support the findings of this study are available from the corresponding authors upon reasonable request.