Although the World Health Organization has declared the end of COVID-19 as a public health emergency [1], the persistence of this disease as a global threat should not be overlooked or underestimated [2]. Vaccination has been regarded as one of the most effective strategies against COVID-19 and reduced global COVID-19 mortality, severe disease, symptomatic cases, and COVID-19 infections [2,3]. Furthermore, studies have shown that COVID-19 vaccine also had a preventive effect against post–COVID-19 condition [4-6].

Despite significant progress made with vaccination efforts, achieving high vaccination coverage remains a challenge due to disparities in vaccine distribution and vaccine hesitancy [7-9]. Disparities in vaccine distribution have been observed between different countries, with vaccination rates varying markedly between high- and low-income countries [10]. In addition, COVID-19 vaccine hesitancy has been reported across countries [11], and booster hesitancy has also become a growing concern for public health officials [12]. Vaccine hesitancy can change over time and in response to different circumstances. Notably, vaccine hesitancy tends to increase when population-level side-effect studies are released after emergency approvals [13]. These challenges underline the need for well-designed vaccination programs to ensure equitable access and high uptake.

Designing a successful vaccination program, including vaccine selection, rollout, and accessibility, is crucial [14,15]. A thorough understanding of individual needs and preferences will allow us to better tailor vaccination programs, which will facilitate the appeal and uptake of COVID-19 vaccines [16,17]. One approach increasingly used to elicit preferences for vaccines and vaccination programs is the discrete choice experiment (DCE) [18,19]. DCEs are scientific research methods that assess preferences by presenting respondents with a series of hypothetical scenarios. In these scenarios, individuals choose among different alternatives which are characterized by specific attributes. By analyzing these choices, researchers can identify the relative importance of each attribute and estimate utility functions [20,21]. DCEs provide valuable insights into decision-making processes and allow for objective evaluation of attribute-based benefits [22-24]. Published studies have been conducted to identify and review choice-based experiments that assess vaccine preferences [18,19]. However, it is important to note that the nature of various vaccines is different, and the preference for vaccines of COVID-19 was not specifically included in these studies.

The COVID-19 vaccines were developed under emergency conditions where there were no peer-reviewed systematic reviews of DCEs on COVID-19 vaccine preference data to inform global decision-making. The diversity in COVID-19 vaccine preferences may be attributed to disparities in vaccine development and production, vaccination scheduling and management, public trust and uptake, as well as vaccine prioritization strategies across various countries and regions [25]. Moreover, new mutant variants are more likely to infect new individuals, highlighting the need for more effective booster vaccines [26,27]. This study provides empirical evidence on the development, implementation, and follow-up of the COVID-19 vaccine and provides references for vaccine decision-making of other infectious diseases.

A literature search was conducted in PubMed, Embase, Web of Science, Scopus, and CINAHL Plus platforms in April 2023. Search terms included discrete choice experiments, COVID-19, and vaccines and related synonyms. Further details are provided in Multimedia Appendix 2.

The inclusion and exclusion criteria are detailed in Textbox 1.

Two reviewers (YH and SF) independently performed a 2-stage screening process to identify eligible studies. In the first stage, titles and abstracts were screened to exclude irrelevant studies using the web-based tool Rayyan (Rayyan Systems, Inc [29]). In the second stage, full-text versions of selected papers were assessed to ensure that the inclusion criteria were met. Both reviewers compared the selected papers at each stage to ensure agreement. Any discrepancy or uncertainty between the reviewers was addressed through discussion until a consensus was reached. If not, a third (senior) reviewer (HJ) was consulted to resolve the disagreement.

The extracted data were recorded and managed in Microsoft Excel (Microsoft Corp) software. Full texts were extracted and reviewed independently by 2 authors (YH and YZ), and any disagreements were resolved by a third reviewer (HJ). Data extraction was performed for 3 specific aspects, focusing on their relevance and importance for the analysis of the DCE: (1) study information (author, publication year, study period, country, population, and sample size); (2) information on the DCE methodology (survey administration, attribute and level selection, pilot-tested, experimental study design, choice sets per respondent, options per choice set, inclusion of an opt-out option, and statistical models); and (3) information on the DCE results (number of attributes, included attributes classified into 4 categories [outcome, process, cost, and other], and the most important attribute).

Choice-based experiments use different definitions for similar attributes [19]. To address this issue, the attributes were initially grouped into 4 main categories: outcomes, process, cost, and other. The outcomes category encompassed the outcomes or consequences of vaccine administration, such as safety and effectiveness. The process category included activities related to the delivery and administration of vaccines, such as service delivery, dosing, and visits. The cost category focused on the financial aspects of vaccines. Any attributes that did not fit into these 3 categories were classified as other, such as disease risk, incentives or penalties for vaccination, vaccine advice or support, and so on. The classification of outcome, process, cost, and other attributes depended on the aim and design of the studies. It should be noted that vaccine effectiveness and safety were phrased differently in different studies. To facilitate a comparison between studies, efficacy [11,30-41], protection rate [42,43], and decreased deaths [44] were summarized as vaccine effectiveness, whereas side effects [11,26,31,35,37,40,41,43,45-61], rare but serious risks [62], and the likelihood of having a flare [62] were summarized as vaccine safety (Multimedia Appendix 3 [11,26,30-74]).

High-income countries (HICs) and low- and middle-income countries (LMICs) were classified according to the World Bank [75]. LMICs encompass low-income, lower-middle–income, and upper-middle–income countries. On the basis of previous literatures [63,76,77], we hypothesized that individuals’ preferences for vaccines may vary depending on the status of the pandemic. Therefore, we sought to explore how COVID-19 vaccine preferences differed during different study periods. To do this, we used data from the surveillance website [78] to define the pandemic periods based on daily COVID-19 cases. The first group, before the pandemic wave, referred to the period before the outbreak of the pandemic, when the number of incident cases was low. The second group, during the pandemic wave, represented the peak of the pandemic or was characterized by a rapid increase in the number of incident cases. The third group, after the pandemic wave, was when the number of incident cases decreased and remained low (Multimedia Appendix 4 [11,26,30-74]).

The 5-item Purpose, Respondents, Explanation, Findings, and Significance (PREFS) checklist, developed by Joy et al [79], is widely accepted and used to assess the reporting quality of preference studies [18,80-84]. It evaluates studies based on criteria such as the study’s purpose, respondent sampling, explanation of assessment methods, inclusion of complete response sets in the findings, and use of significance testing.

This review used a combination of text and summary tables to effectively convey information about the characteristics and results of the included studies. Descriptive statistics were used to summarize the study characteristics. The findings were synthesized in a narrative format, providing an overview of the included studies, highlighting the key features of the study designs, and presenting the main findings of the COVID-19 vaccine preference studies. Subgroup analyses were performed by independent factors such as HICs or LMICs and study period (before, during, and after the pandemic wave).

The search yielded a total of 623 records. After title and abstract screening, 513 (82.3%) records were excluded. An additional 63 (10.1%) studies were excluded after full-text assessment. Finally, 47 (7.5%) studies met the eligibility criteria and were included in the review (Figure 1).

We included 47 studies from 29 countries. Among them, 5 (11%) studies were conducted in multiple countries, with 4 studies conducted in both HICs and LMICs and 1 study conducted in >1 HICs. In addition, 22 (47%) studies were conducted in HICs, while 21 (45%) studies were conducted in LMICs. China stood out with the highest number of preference-based DCEs for COVID-19 vaccines, with 19 (40%) studies. The United States followed closely with 9 (19%) studies, followed by France (n=5, 11%), the United Kingdom (n=4, 9%), Germany (n=4, 9%), and Spain (n=3, 6%). Australia, Canada, India, Italy, Japan, the Netherlands, and South Africa had 2 (4%) studies each. All other countries had only 1 (2%) study (Figure 2). The studies were published between the years 2020 and 2023, with sample sizes ranging from 194 to 13,128 participants. The median number of participants per study was 1456 (IQR 872-2109).

Most participants were adults, although the specific focus varied. Most studies (36/47, 77%) involved general population samples, whereas some studies (11/47, 23%) included specific groups of participants. These included 5 studies conducted in universities using web-based tools, including 3 studies with university students and 2 studies with both students and staff. In addition, 3 studies involved health care workers (Chinese intensive care unit clinicians, health care workers, and health care and welfare workers); 2 studies involved parents with children aged <18 years, and 1 study involved people with chronic immune-mediated inflammatory diseases (Table 1).

Among these 47 studies, researchers commonly used a multifaceted approach to identify and select attributes and levels. Among the studies reviewed, 23 (49%) studies reported a literature review with qualitative assessments such as expert interviews and public surveys. A total of 25 (53%) studies reported a pilot DCE survey. In terms of survey administration, most studies (40/47, 85%) reported that the DCE was conducted through web-based surveys (Table 2).

Of the 286 attributes identified in the 47 studies, 126 (44.1%) were categorized as outcome attributes, followed by 82 (28.7%) as process attributes, and 22 (7.7%) as cost attributes. The remaining 55 (19.2%) attributes were categorized as other attributes (Table 3 and Multimedia Appendix 3).

In total, 2 of the 5 multicountry studies did not report preferences for each country and were therefore excluded from the synthesis of the most important attribute. A total of 53 data points on COVID-19 vaccine preferences were collected from the study population of the corresponding country. In the outcome category, among the 30 attributes examined, effectiveness emerged as the most prominent, accounting for 40% (21/53) of the studies [31,35,36,38-42,48,50-52,57,58,60-62,64-67]. Safety was addressed in 13% (7/53) of the studies [33,43,47,56,59,68,69], while protection duration was mentioned in 4% (2/53) [11,50]. In the process category, 13 attributes were identified. Brand (1/53, 2%) [32], region of vaccine manufacturer (1/53, 2%) [34], and halal content (1/53, 2%) [53] were associated with vaccine production. In addition, waiting time for COVID-19 vaccination (1/53, 2%) [70] and vaccine frequency (1/53, 2%) [71] were considered. Furthermore, 3 (6%) studies on vaccine distribution prioritized vaccination for the medical risk group (1/53, 2%) [72], those who had a higher COVID-19 mortality risk (6/53, 11%) [63], and those who had the potential capacity to spread the virus (1/53, 2%) [72]. In the cost category, personal vaccination cost accounted for 6% (3/53) [31,37,41]. Among the other attributes (7/53, 13%), disease risk threat was of particular importance, including possible trends of the epidemic (1/53, 2%) [30] and COVID-19 mortality rate (1/53, 2%) [55]. In addition, incentives and penalties for vaccination were identified, including quarantine-free travel (1/53, 2%) [33] and mandatory testing at own expense if not vaccinated (1/53, 2%) [44]. Vaccine advice or support included vaccination invitation sender (1/53, 2%) [73] and recommenders (1/53, 2%) [46]. The proportion of friends and family members who had received the vaccine (1/53, 2%) [26] was also among the other attributes influencing decision-making (Table 2).

Although effectiveness remained the most important attribute, it is worth noting that variations in preferences were also observed among different subgroups. A higher proportion of studies conducted in LMICs (4/24, 17%) than in HICs (3/29, 10%) prioritized on safety (Multimedia Appendix 5). In addition, COVID-19 mortality risk was the second most important attribute (6/29, 21%) after effectiveness in HICs. Cost was considered to be another most important attribute (3/24, 13%) in LMICs. Interestingly, many other attributes also became more important as the pandemic progressed. Protection duration (2/24, 8%) emerged as one of the most important attributes during the pandemic wave. COVID-19 mortality risk (5/25, 20%) and cost (3/25, 12%) were considered as the most important attributes after the pandemic wave (Multimedia Appendix 6).

The overall reporting quality was deemed acceptable but there is room for improvement. The PREFS scores of the 47 studies ranged from 2 to 4, with a mean of 3.23 (SD 0.52). No study scored 5. Most studies scored 3 (32/47, 68%) or 4 (13/47, 28%), while 2 studies (2/47, 4%) scored 2 (Multimedia Appendix 7 [11,26,30-74]).

This systematic review synthesizes existing data on preference for COVID-19 vaccine using DCE, with the aim of informing improvements in vaccine coverage and vaccine policy development. We identified 47 studies conducted in 29 countries, including 21 HICs and 8 LMICs. HICs had an adequate supply of vaccine since the early emergency availability of COVID-19 vaccine, and HICs had 1.5 times more doses of COVID-19 vaccinations than LMICs by September 2023 [85]. In total, 19 (40%) studies were conducted in China and 9 (19%) in the United States, demonstrating their significant contribution to the research and their leadership in vaccine research and development. Vaccine effectiveness and safety were the most important attributes in DCEs, although preferences differed among subgroups.

Recent years have seen new trends in the design, implementation, and validation of the DCE. For example, most studies (40/47, 85%) reported that the DCE was administered through web-based surveys, which have become a quick and cost-effective way to collect DCE data [66]. Almost half of the studies (25/47, 53%) did not report a pilot test. However, piloting in multiple stages throughout the development of a DCE is conducive to identifying appropriate and understandable attributes, considering whether participants can effectively evaluate the full profiles, and producing an efficient design [21,86,87].

Overall, vaccine effectiveness and safety have emerged as the most commonly investigated attributes in the outcome category. Despite heterogeneity in preferences across subpopulations, effectiveness remains the primary driver for COVID-19 vaccination across the studies [31,35,36,38-42,48,50,51,57,58,60-62,64-67], similar to the previous findings [18]. A study conducted in India and Europe found that respondents’ preference for the COVID-19 vaccine increased with effectiveness and peaked at 95% effectiveness [45]. Another study conducted among university staff and students in South Africa found that vaccine effectiveness not only was a concern but also significantly influenced vaccine choice behavior [64]. Interestingly, a nationwide stated choice survey in the United States found a strong interaction between effectiveness and other attributes [58]. These findings support the ongoing efforts to maximize vaccine effectiveness while emphasizing the importance of communicating information on vaccine effectiveness to the target population for promotion [62].

Safety has also been identified as a crucial factor influencing the acceptance of COVID-19 vaccine [33,43,47,56,59,68,69]. One study indicated that the likelihood of the general public choosing vaccines with low or moderate side effects increased by 75% and 63%, respectively, compared with vaccines with high side effects. While the likelihood changed within a 30% range when most attributes other than effectiveness and safety were changed [69]. In addition, respondents in Australia expressed a willingness to wait an additional 0.04 and 1.2 months to reduce the incidence of mild and severe adverse events by 1/10,000, respectively [56].

Similar to the results of previous systematic reviews of DCEs for various vaccines [18,19], the most common predictors of COVID-19 vaccine acceptance are effectiveness and safety, particularly during the rapid development and rollout of COVID-19 vaccines, which essentially boils down to trust in the vaccine [31]. Respondents expressed the importance of having a safe and effective COVID-19 vaccine available as soon as possible, but the majority preferred to wait a few months to observe the experience of others rather than be the first in line [43]. Therefore, collaborating to enhance vaccine effectiveness while reducing the risk of severe side effects could be a highly effective strategy to address vaccine hesitancy and augment vaccine desirability. Dissemination of this important vaccine-related information by governments and health care institutions, along with effective communication by health care professionals, can help build public trust and ultimately increase vaccination rates [69]. However, these inherent vaccine attributes are typically beyond the control of a vaccination program, and given the ongoing mutations of SARS-CoV-2, it is challenging to predict the effectiveness of the vaccines currently in development [66]. Global collaboration between scientists and pharmaceutical companies is therefore essential to improve vaccine effectiveness and minimize side effects [41].

Vaccine production, including its origin, brand, vaccine frequency, and content, are key considerations in the process category. Vaccine brand also has a significant impact on vaccine choice [32], independent of effectiveness and safety, due to factors such as reputation, country of origin, technological advances, and reported side effects associated with the brands [35]. For vaccine origin, some studies found that participants preferred domestic vaccines to imported vaccines, which may depend on the availability or the approval of vaccines in different countries [31,41,50] or the incidence of side effects among different types of COVID-19 vaccines [37]. However, some studies found that imported vaccines were more likely to be accepted than domestically produced vaccines, which may be attributed to less trust in domestically produced vaccines [57,66]. A study on vaccine preferences among the Malaysian population found that the composition and production process of the COVID-19 vaccine, which complied with Islamic dietary requirements (ie, halal content) was an important factor for many Malaysians when deciding whether to be vaccinated. This underscores the substantial influence of religion on vaccine choice [53].

Vaccine frequency was emphasized to play an important role in the choice of COVID-19 vaccine among the US public, while the 90% efficacy with low side effect rate of the COVID-19 vaccine was set. The prospect of vaccinating once to get lifelong immunity was very attractive, reflecting the fact that people were effort minimizers [71]. This is similar to the nature of the 2 studies referenced in the outcome attribute, where the protection duration is prioritized. Given the threat of COVID-19, people expect the protection duration to be as long as possible [11,50].

When vaccine supply is limited, people tend to prioritize vaccination for those who are more susceptible to the disease, have higher mortality rates from infectious diseases, or have greater potential to spread the virus. A study in Iran found that individuals tend to prioritize vaccination for those in the community with higher potential for virus transmission [57]. In addition, results from a study in 6 European countries revealed unanimous agreement among respondents that candidates with higher mortality and infection risks should be prioritized for vaccination [63]. While another study conducted among Belgians also found that respondents would prioritize populations at higher medical risk [72].

Cost was another important factor influencing COVID-19 vaccine preferences, mostly related to out-of-pocket costs [31,37,41]. In 2 studies comparing public preferences for COVID-19 vaccines in China and the United States, vaccine efficacy emerged as the most important driver for the American public, whereas the cost of vaccination had the greatest impact on the Chinese public. This difference was likely due to the relatively stable pandemic situation in China at the time and the lower perceived risk of COVID-19. As a result, the Chinese population was more price sensitive and reluctant to pay for vaccination [31,37,41].

For the other category, several different attributes were highlighted, depending on the specific population or situation. When people perceive the threat of a disease, their desire to be vaccinated becomes more urgent. In a study among health care workers in China, participants’ expectations about the future development of COVID-19 had a greater impact on their decision to be vaccinated than their perceived risk of infection or actual case rates, which may have been influenced by their previous experience with seasonal influenza vaccination [30]. The mortality rate of COVID-19 was considered the most influential factor in the uptake of COVID-19 booster shots in Vietnam. This study was conducted during a pandemic wave in Vietnam, which may have led to an increased perception of public health risks and a greater inclination toward COVID-19 vaccination [55]. To achieve herd immunity, government authorities can implement policies of incentives and penalties for vaccination to encourage population-wide uptake. A study conducted in the Netherlands revealed that respondents particularly disliked policies that penalized those who were not vaccinated, such as mandatory testing at their own expense if they were not vaccinated [44]. Instead, they favored policies that rewarded vaccination, such as giving vaccinated individuals additional privileges through a vaccination passport. This finding is consistent with a study in Hong Kong, which found that quarantine-free travel was considered the most important motivator among university students and staff, given their frequent engagement in international travel [33].

The source of vaccine information also influences vaccine decision-making [30]. Variation in the sender of vaccination appointment invitation via SMS text messaging and recommenders may potentially influence the public’s willingness to vaccinate against a disease [30,46,73]. Furthermore, the acceptance of vaccines was observed to change as the firsthand information about vaccine side effects and effectiveness was provided by friends and family in India [26].

In HICs, COVID-19 mortality risk was the second most important attribute after effectiveness, as respondents in all 6 high-income European countries from a study of public preferences for COVID-19 vaccine distribution prioritized candidates with higher mortality risks [63]. However, individuals from LMICs appeared to be more concerned about vaccine safety than those from HICs. This may be related to greater confidence in vaccine safety in HICs due to the earlier initiation and higher rates of COVID-19 vaccination [85]. In contrast, in some LMICs, vaccine safety was reported as the main reason influencing the willingness to vaccinate due to the rapid development of the COVID-19 vaccines [26,43,47,59,68,69,74,88].

Interestingly, the preference for COVID-19 vaccines may also have changed as the pandemic progressed [63]. Similarly, effectiveness remained the most important attribute in all periods, possibly due to the continuing severity of the pandemic and the fear of the possible emergence of new coronavirus strains [43]. Before the pandemic wave, the information on vaccine effectiveness was limited [26], but people still considered vaccine effectiveness to be the most important driver of vaccination. However, during the pandemic, the public’s perception of the health risk increased. As vaccines were introduced and used, people seemed to become more concerned about the duration of vaccine protection and preferred a longer vaccine protection [11,50]. After the pandemic wave, as the pandemic situation gradually stabilized, cost, combined with their perception of the risk of susceptibility, became more important in their preferences. However, despite this shift, most of the public still believed that people who are at higher risk of infection or death should be vaccinated first [63].

Our study had several limitations. First, not all studies used the same attributes and levels, which limited our ability to perform a quantitative synthesis and directly compare the estimates of model parameters. Instead, we qualitatively synthesized and summarized the range of attributes that may be useful in the formative stage of attribute selection in future DCE surveys investigating the preference for COVID-19 vaccine. Second, although DCEs have been shown to be a valid method for eliciting preferences, the experiment may not represent real market choices but rather hypothetical scenarios with plausible and realistic attributes. However, it offers opportunities to evaluate vaccines that are not yet available in the market or to specific population [68]. Third, the commonly used classification of outcome, cost, and process was used in order to better explain the public’s preference for vaccine attributes. However, several attributes could not be properly classified, and a fourth category (ie, other attributes) had to be added [19]. Meanwhile, the variety of attributes included may make it difficult to appropriately name and interpret this category as a whole. Fifth, the PREFS checklist is limited to 5 questions and fails to elicit several criteria that should be reported in DCE studies. Also, it does not provide sufficient tools to assess the biases in a DCE, such as selection bias and nonresponse bias [79,89]. Finally, although there was no specific theoretical framework to structure our qualitative analysis from the 4 identified categories, our classification was based on previous studies [18,19,82,90,91] and our own findings. This synthesis led us to categorize attributes into 4 main classes, providing a clear structure for analyzing and presenting participants’ vaccine preferences and making it easier to compare their preferences across different studies.

In conclusion, this systematic review synthesized the global evidence on preferences for COVID-19 vaccines using the DCE methodology. Vaccine effectiveness and safety were found to be the main drivers for COVID-19 vaccination, highlighting the importance of global collaboration to improve vaccine effectiveness and minimize side effects, as well as the importance of communicating this vaccine-related information to the public to maximize the uptake of COVID-19 vaccines. The subgroup analyses emphasized the importance of differences in vaccine preference of specific populations and time periods in optimizing the acceptance of COVID-19 vaccines. These findings may serve as valuable insights for government agencies involved in the social mobilization process for COVID-19 vaccination. However, the response to the pandemic is a continuous learning process [92]. It is crucial for policy makers to consider preference evidence when designing policies to promote vaccination.