The adapted COM-B model has three theoretical constructs: (1) capability, (2) motivation, and (3) opportunity. The WHO Regional Office for Europe also provided examples for each construct (see Table 1). Capability refers to the individual’s physical and psychological ability to perform the behavior, along with the knowledge and skills required to complete the activity [17]. In particular, psychological ability is the ability of an individual to have the necessary thought processes, such as being able to understand, reason, etc [17]. Motivation has a broad definition that includes goals and conscious decision-making as well as all other individual processes that motivate and lead to behavior, such as automatic processes (habitual processes, emotional responses) and reflective processes (analytical decision-making) [17]. Opportunity refers to the contextual factors that prompt the behavior to occur, including the physical opportunity and the environment and social opportunity [17]. The adapted COM-B model also provides a dynamic relationship between constructs. For example, both capability and opportunity can affect motivation. All three constructs of competence, motivation, and opportunity can generate behavior; on the other hand, behavior can influence these three constructs.

We collected English tweets posted from November 1-22, 2020, using a combination of relevant keywords and hashtags: (#covid OR covid OR #covid19 OR covid19) AND (#vaccine OR vaccine OR #vacine OR vacine OR vaccinate OR immunization OR immune OR vax). After excluding retweets, we randomly selected 5000 tweets for manual coding and content analysis. The random numbers were generated through the NumPy package in Python. Then, we mapped the random numbers with the index of collected tweets.

We performed a content analysis informed by the adapted COM-B model. The coding schema was developed iteratively. First, we developed the coding schema based on the definitions of constructs in the adapted COM-B model. Two reviewers (SL and JL) independently coded 1000 tweets in each round. After completing one round of coding, the two reviewers met with a third reviewer to discuss disagreements and update the coding schema until consensus was reached. We calculated the interrater reliability for the last round. If a tweet mentioned ≥2 constructs simultaneously, we coded it with multiple labels. We conducted chi-square tests to explore the relationship between theoretical constructs with the positive/negative behavioral intention. The statistical significance threshold was .05.

We coded 5000 COVID-19 vaccine–related tweets, which were posted by 4796 unique users. We found 279 tweets that stated their behavioral intentions. The remaining tweets did not state any behavioral intentions toward COVID-19 vaccines. A total of 97 tweets were labeled with positive behavioral intentions, while 182 tweets contained negative behavioral intentions. Among them, 28 tweets were mapped with capability factors; 155 tweets were related to motivation; 23 tweets were related to opportunities; and 74 tweets did not contain any useful information about reasons for their behavioral intentions (see Figure 2). The κ value was 0.73. Of the tweets, 2 mentioned ≥2 constructs at the same time. Tweets that mentioned capability (χ²₁=17.286, P<.001), motivation (χ²₁=35.558, P<.001), and opportunity (χ²₁=4.545, P=.03) were more likely to have negative behavioral intentions (Table 2).

The first theme regarding capability was that some users lacked knowledge about the COVID-19 vaccines and were influenced by misinformation. mRNA vaccines are a new technology, and some believe that these vaccines can alter DNA. As for the vaccine itself, some users believe that it contains a microchip that can be tracked by the government. Some people expressed that they did not know the side effects of COVID-19 vaccines or even their long-term effects. Based on this concern, misinformation was generated that the vaccine could cause sterility, cancer, etc. Other reasons for not getting the vaccine included the belief that the vaccine could make people sick, the belief that there is no need for people who have been infected with COVID-19 to get the vaccine, and that the vaccine does not prevent infection but only alleviates symptoms. All these beliefs reflect a lack of understanding of the COVID-19 vaccines among users. Having more knowledge about vaccines could help people develop positive behavioral intentions; for example, some people mentioned that they could understand why the development process was fast because researchers only needed to isolate the correct strain rather than create one. Others mentioned understanding that vaccines do not contain live viruses, so they would be willing to get the vaccine.

Some users emphasized that their physical condition was not suitable for COVID-19 vaccination and that they were unsure how their body would react after vaccination (eg, recently had an influenza vaccine shot, had a suppressed immune system, had a severe adverse reaction to vaccination in the past, had a stroke).

In the physical opportunities category, some users said they would not accept the vaccine if they had to pay for it. Many users said vaccination should be a free choice and that they would refuse to receive the vaccine if it became mandatory. Some users wanted more regulation of the COVID-19 vaccine, and some were concerned about the availability of information.

Among the social opportunities, we found some factors influencing the COVID-19 vaccination related to social and cultural demands and support, such as going against religion, defying culture, health workers not recommending vaccination, and family members actively discouraging vaccination. Others intended to be vaccinated because of the fear of the social consequences and reactions to vaccination, such as fear of affecting their work and requiring proof of vaccination for many activities in the future.

Most of the reasons for behavioral intentions were categorized into the motivation construct. Many users expressed attitudes and opinions about the COVID-19 vaccine or the disease itself, such as not considering the disease to be severe or life-threatening and not considering the vaccine to be effective (because of low efficiency and mutation of the virus). Others assessed the risks and considered the rushed vaccine to be more harmful than COVID-19. Some of those with positive behavioral intentions stated that they chose vaccination because they did not want to be infected with COVID-19.

Some users expressed strong emotions and feelings about COVID-19 vaccines, such as fear of the vaccine and concern about its side effects because it is new. Notably, others felt that vaccination was selfish or unfair, as they would not be exposing themselves to the same risk as others who survived COVID-19; thus, they were reluctant to be vaccinated.

Other themes were values and beliefs. Some users expressed a belief that the body’s natural immune system is better than any vaccine, or they believed more in alternative medicine. On the other hand, others who wanted to be vaccinated emphasized the positive values of vaccination, such as saving more lives, reopening the economy, and returning to normal life.

Confidence and trust were dominant themes. Users expressed distrust in many areas: the quality of the vaccines (hastily manufactured, untested), health authorities, science, companies, and government. For Twitter users in the United States, we found that unlike with other vaccines, part of the reason people do not trust the government is because of their past handling of the COVID-19 pandemic. In addition, the COVID-19 vaccine rollout came during a time when a new president was being elected in the United States, and some people lacked confidence in the newly elected president’s party.

In this study, we conducted a theory-based content analysis using a dataset of 5000 tweets posted from November 1-22, 2020. We identified 279 tweets that contained behavioral intentions regarding COVID-19 vaccines and mapped them to constructs in the adapted COM-B model. We generated nine themes that influence Twitter users’ intentions to receive COVID-19 vaccines. The constructs in the COM-B model could be applied systematically to characterize factors that influence behavioral intentions toward COVID-19 vaccines. In addition, we found that among tweets that simply stated behavioral intentions without including any reason, the number of positive intention tweets was higher than that of negative intention tweets. The results also implied that more than half of the tweets expressing the decision-making process were negative intention tweets. This finding aligns with our expectation of understanding the factors that contribute to vaccine hesitancy to better develop tailored vaccine promotion programs.

The novelty of COVID-19 vaccines and the current social context have created further difficulties in vaccine rollout. Identified barriers of influenza vaccination intention and behavior include lack of confidence (eg, negative attitudes, mistrust), inconvenience (eg, cost, access), calculation (eg, risk assessment), and complacency (eg, underestimating disease severity) [26]. Our results revealed the presence of several other factors that influence behavior toward COVID-19 vaccination. First, misinformation or conspiracy theories about COVID-19 and COVID-19 vaccines are much more prevalent on social media than those about other diseases or their related vaccines [27-29]. Some users were influenced by this misinformation, and this led to refusal of the vaccine. Second, Twitter users expressed concerns about mandatory vaccination. At the Emergency Use Authorization stage, mandatory vaccination is legally and ethically questionable [30]. However, with full Biologics License Application approval, policy makers may mandate vaccination for all populations. Given the existence of users who have strongly indicated that they would not accept mandatory vaccination against COVID-19, policy makers must be cautious in determining vaccination policies for the public. Some studies have suggested that vaccine mandates do not improve vaccine acceptance rates, and a proposed alternative approach is to apply informed risk communication and give people the freedom to choose without compromising personal autonomy [30]. Third, users with positive behavioral intentions emphasized that the positive value of vaccination to society, such as restoring economic and normal life from before the epidemic, motivated them to be vaccinated. This facilitator is uncommon in other vaccination behaviors. This facilitator could be matched with a strategy of converting personal decisions into public acts [31]. Fourth, we observed that some users were reluctant to be vaccinated because they felt that the COVID-19 vaccine was unfair to those who survived. Based on this concern, it may be helpful to select COVID-19 survivors as opinion leaders to promote COVID-19 vaccines, increasing public awareness of the severity and risk of the disease. Fifth, in addition to the mistrust of vaccines and science that also exists for other vaccines, for the COVID-19 vaccine, users in the United States expressed more mistrust of the government for two specific reasons: (1) the previous administration's inappropriate behavior in handling COVID-19 and (2) the lack of confidence in the newly inaugurated president’s political party. These findings were also aligned with previous studies that proposed a role of politics in COVID-19 vaccine hesitancy [8]. Sixth, it is worth noting that even though previous studies have identified that past vaccination behavior can be used to predict future vaccination behavior [32], we found that past experiences with other vaccines may not affect COVID-19 vaccination. For example, one user mentioned:

The above differences contribute to the fact that the rollout of COVID-19 vaccines could be more complicated than that of other vaccines. Researchers need to develop interventions specific to the COVID-19 vaccine to improve acceptance rates. This also provides an opportunity for future studies to comprehensively analyze why behavioral intentions toward COVID-19 vaccines are different from those toward other vaccines.

In several studies, Twitter content analyses have been conducted of health care behaviors other than COVID-19 vaccination. For example, Chew and Eysenbach [14] collected 2009 H1N1-related tweets and identified the resource content posted most often, followed by personal experience, personal opinion, jokes, marketing, and spam. Li et al [33] extracted COVID-19 stigma–related tweets and found that group labeling, responsibility, and peril tweets disseminated the stigma. Furthermore, several studies have validated the usability of tweets through theory-based content analysis to promote breast cancer promotion programs [16,34]. Our study is the first study to analyze the behavior intention of COVID-19 vaccines through a theory-based content analysis using social media content.

This study has several limitations. First, we only analyzed the behavioral intentions of users on Twitter. Previous studies have shown that health care providers are the primary advocates for vaccination and largely influence vaccination acceptance rates [35-38]. In this platform, we could not distinguish users’ occupations. However, the themes reported in this study could help researchers to develop evidence-based interventions for the general public. To examine the vaccine behavior of health care providers, we will conduct a questionnaire-based survey, and the results from that study could aid the development of clinical guidelines for health workers. This approach of considering the general public separately from health care providers is also recommended by the TIP approach developed by the WHO Regional Office for Europe [24]. Second, there are differences between behavioral intentions and actual vaccine behaviors. However, behavioral intentions have been shown to directly influence actual behaviors [39-41]. Third, Twitter users are considered to be younger and have a higher level of education than the general public [13]. Based on this concern, further qualitative research could be conducted on the older population, those with lower education levels, or those with limited access to the internet.

In future research, a literature review could be conducted to summarize current implementation strategies for COVID-19 vaccine promotion and map them to the themes identified in this study to determine gaps in recent research. The inner mechanism of the adapted COM-B model—the behavior change wheel—could inform evidence-based and theoretical implementation strategies to improve the effectiveness of COVID-19 vaccine promotion programs.

The study demonstrates the capability of applying the COM-B model to characterize behavioral intentions toward COVID-19 vaccines on the Twitter platform. We successfully generated nine themes of factors that affect behavioral intentions. Positive behavioral intentions were affected by the positive values of vaccination (eg, reduced risk of infection, socioeconomic recovery, return to normal life). In contrast, negative behavioral intentions were associated with attitudes and perceptions about COVID-19 vaccines or the disease itself (eg, underestimation of disease severity, low vaccine effectiveness), values and beliefs (eg, greater belief in the natural immune system), confidence and trust (eg, distrust of government or vaccines), and lack of knowledge. The generated themes could be used to create theory-based and evidence-based implementation strategies to promote COVID-19 vaccines.