Data were collected using an online questionnaire from November 26 to December 8, 2021, by an external provider of online surveys (Bilendi GmbH). This period coincided with increased media coverage of the first booster vaccinations for COVID-19. The sample was a quota sample, representative of age, gender, and place of residence distribution within a single federal state according to Eurostat 2018. Bilendi invited registered users to participate, offering remuneration for completion. The invitation included information on survey duration, compensation, and a link to the questionnaire. Eligible participants were German-speaking individuals aged 18-75 years.

The questionnaire was divided into three sections: (1) internet use for COVID-19 vaccination information, knowledge about the vaccination, and health literacy; (2) attitudes toward vaccination, vaccination behavior, and determinants of vaccine willingness; and (3) sociodemographic characteristics.

This research project was approved by the Ethics Committee of Osnabrück University (reference Ethik-46/2021 and Ethik-68/2021). Bilendi ensured data protection and anonymity in accordance with the German Data Protection Act. The participants were informed about the method of data collection, anonymized data transfer, and compensation during panel registration, and were assured that their responses would remain anonymous and not be shared with third parties. Participation was voluntary and could be terminated at any time without penalty. The participants accessed the survey at Respondi Mingle [21] after accepting the terms and conditions and data protection guidelines.

The frequency of using the internet to search for information regarding COVID-19 vaccination was recorded using the following items: daily, several times a week, once a week, less than once a week, not at all, and use of other sources.

Channels used to search for information: the participants were asked about the most common online information channels. The responses were initially coded as follows: perceived fleetingly, consciously searched for information, no contact, and do not use. Since the options “no contact” and “I don’t use” were similar, these were combined into “consciously searched for information,” “fleetingly perceived,” and “no contact/I don’t use.”

Formats: the participants were asked which formats related to COVID-19 vaccination they encountered online (text articles, audio articles, videos, posts on social networks, posts via messenger services, discussion forums), using a scale of frequently, sometimes, rarely, and never.

Reasons for information-seeking: the reasons for seeking information about COVID-19 vaccination online included obtaining in-depth information about the vaccination, researching information, or exchanging views with like-minded individuals about the benefits and risks of the vaccination. The participants confirmed these statements with “yes” or rejected them with “no.”

Age was categorized into 4 groups: 18-29 years, 30-44 years, 45-64 years, and 65-75 years. Education was recorded based on school education and professional qualifications according to the demographic standard of the Federal Statistical Office of Germany [22]. Education levels were then clustered into basic, medium, and higher educational attainment using the Comparative Analysis of Social Mobility in Industrial Nations classification [23]. Immigration history was assessed according to the recommended standards in Germany [24,25]. A distinction was made between two categories: (1) immigrants and their (direct) descendants (ie, participants not born in Germany or participants born in Germany with both parents born outside of Germany) and (2) individuals without an immigration history (ie, participants and both parents born in Germany) [26]. Place of residence was categorized into rural areas (up to 5000 inhabitants), small towns (5001 to under 20,000 inhabitants), medium-sized cities (20,000 to 100,000 inhabitants), and large cities (more than 100,000 inhabitants) [27].

A total of 6 items on general health literacy were developed based on Schaeffer et al’s [28] study. Participants answered the following questions: When you think back to the time before the coronavirus pandemic, how easy was it for you to (1) obtain health-related information, (2) comprehend health-related information, (3) evaluate the content of health-related information, (4) assess the credibility of health-related information, (5) critically question health-related information, and (6) use health-related information as a basis for decision-making (translation from German). The answer options from the Likert scale were coded as follows: 1=very difficult, 2=difficult, 3=neither, 4=easy, 5=very easy. By summing the results, a new variable, “health literacy score,” was created (range from 6 [minimum] to 30 [maximum]). The interitem correlation for this survey is α=.86.

For the knowledge question on COVID-19 vaccination, 5 statements were developed based on information or vaccination myths published on the Bundesministerium für Gesundheit website [29].

Please evaluate the following statements: (1) An mRNA vaccine can alter human DNA. (2) Vaccine reactions such as pain at the injection site, fatigue, headache, and muscle pain are signs that the immune system is responding to the vaccine as intended. (3) Depending on the vaccine, vaccinated individuals are 65%-95% less likely to contract COVID-19 than those who are not vaccinated. (4) After vaccination, I no longer have to follow the rules of conduct (keeping distance, wearing a mask). (5) Full immunity is achieved immediately after vaccination.

A total score (knowledge score from 0 to a maximum of 5 points) was calculated with values for each item as either “not known” (0) or “known” (1).

The participants were asked whether they felt external pressure to be vaccinated, for example, from their employer or the general population, or whether they felt free to decide for or against vaccination (1 item, possible answers: yes/no).

Self-efficacy was measured using the current version of the validated General Self-Efficacy Scale by Jerusalem and Schwarzer [30]. The self-assessment procedure comprises 10 items and is used to record general optimistic self-beliefs. The interitem correlation for this survey is α=.92.

The validated 7C Vaccination Readiness Scale by Geiger et al [7] was used to assess readiness to vaccinate, evaluated as recommended by the authors. The scale contains 7 components of readiness to vaccinate: confidence, complacency, constraints, calculation, collective responsibility, compliance, and conspiracy, each with 3 associated items (range: 1=strongly disagree to 7=strongly agree). The instrument is publicly available [31]. Readiness to vaccinate is represented as the average value across all items of the 7C scale. The internal consistency of the scale was examined using McDonald’s omega (ω). The value of ω=0.96 indicates very good internal consistency.

For vaccination status, it was first determined whether the individuals had received 1 or 2 vaccine doses (from BioNTech/Pfizer, Moderna, or AstraZeneca) or 1 vaccine dose from Johnson & Johnson or none to date, which corresponded to the vaccination recommendation in Germany. This information was then summarized into 2 categories—vaccinated or unvaccinated—to represent the participants’ vaccination status.

Analyses on online information behavior were only conducted with a group of 889 participants who searched online for information about the COVID-19 vaccination. Absolute and relative frequencies were calculated for the formats and channels used, the reasons for searching for information about COVID-19, and the frequency of searching for information on the internet over the last 4 weeks from completing the questionnaire. A cluster analysis was conducted to determine user groups regarding information behavior about COVID-19 vaccination online. The channels, formats, frequency of information behavior, and reasons for searching for information about the vaccination were used as potential variables for determining user groups through a cluster analysis.

Before the cluster analysis, a Spearman correlation analysis was conducted between the ordinal-scaled variables. Significant, strong correlations were identified between the variables channels and formats (podcast channel and audio contributions: Spearman ρ=0.58, P<.001; WhatsApp and posts via messenger services: Spearman ρ=0.67, P<.001; Facebook and posts in social networks: Spearman ρ=0.63; P<.001). Consequently, the format variables were excluded, which can also be justified because they lacked variation. For the cluster analysis, a hierarchical cluster analysis using complete linkage and Gower distance was applied in RStudio (Posit PBS). After removing cases with missing values and outliers (identified through single-linkage analysis) from the group of 889 participants searching for information online, 778 cases were included in the cluster analysis. The optimal number of clusters was determined using the elbow method and the silhouette method, both of which indicated that 2- and 3-cluster solutions were appropriate. Based on content-related considerations, the 3-cluster solution was selected. A subsequent discriminant analysis showed that 86.4% (672/778) of the originally grouped cases were correctly classified, indicating high quality and accuracy of the cluster analysis and supporting the decision to adopt the 3-cluster model.

Bivariate analyses (chi-square test and ANOVA) were conducted to identify differences between user groups. To analyze the correlations between the user groups and personal characteristics, a binary logistic regression was performed with 3 models for each user group. Model 1 included sociodemographic characteristics (gender, age, education, immigration history, and size of place of residence). Model 2 added willingness to vaccinate, and model 3 incorporated knowledge about vaccination and health literacy.

Additionally, a binary logistic regression analysis was conducted to predict vaccination status. In the first model, the correlation with user groups was tested. The second model included sociodemographic variables, while the third model added potential influencing factors such as vaccination readiness score, knowledge, health literacy, external pressure, and self-efficacy expectation. These analyses were performed using IBM SPSS Statistics, version 29.

In total, 1000 people living in Germany completed the questionnaire. Among the participants, 50% (n=500) identified themselves as male and 50% (n=500) as female. The average age was 46.4 (SD 15.5) years, with ages ranging from 18 to 75 years. Further sociodemographic characteristics of the study population revealed that 8.6% (n=86) had a history of immigration. In terms of educational attainment, 12.9% (n=129) had a basic level, 57.8% (n=578) had a medium level, and 27.1% (n=271) had a higher level. More than half of the respondents lived in a medium-sized town or a large city. Overall, 86% (n=860) of the participants had been vaccinated at least once, while 13.3% (n=133) remained unvaccinated at the time of the survey. Attitudes toward vaccination were as follows: 13.7% (n=137) were (somewhat) opposed, 6.9% (n=69) were undecided, and 78.7% (n=787) were (somewhat) in favor. An overview is provided in Table 1.

In the past 4 weeks, most respondents (N=1000) reported searching the internet for information on COVID-19 vaccination daily (n=266, 26.6%) or several times a week (n=356, 35.6%). Conversely, 11% (n=110) did not search online at all for the information or used other sources.

Further analyses of information behavior were conducted only with the subsample of 889 online users. Websites of health authorities (n=405, 45.6% consciously searched) and online news sites (n=275, 30.9% consciously searched) were used more frequently than health insurance company websites (80/889, 9%). Social media and messenger services were used less intentionally for information gathering. Facebook, WhatsApp, and Instagram were perceived as less reliable sources, while Telegram and Twitter were used by only a few online users (Figure 1).

Text content was the most frequently selected method for learning about COVID-19 vaccination (405/889, 45.6% often and 312/889, 35.1% sometimes). Posts on social media and audio and video content were used less frequently, while discussion forums and messages via messenger services were rarely selected (Multimedia Appendix 1).

Of the 889 online users, 85.9% (n=764) stated that their reason for searching for information on COVID-19 vaccination was to stay updated on new developments. Additionally, 59.4% (n=528) sought to learn more about the vaccination, and 69.4% (n=617) wanted to verify information and research claims. A less substantial reason for using the internet was the intention to persuade others of their views on vaccination through social media posts (131/889, 15.1%). Furthermore, 43.6% (388/889) of respondents reported regularly discussing the advantages and disadvantages of vaccination with others, and half sought information outside of public institutions.

A logistic regression was performed to identify the factors influencing vaccination status (Table 3). Model 1 indicated that the user group is not related to vaccination status. When personal characteristics were added in model 2, education and place of residence emerged as significant influencing factors. Model 3 demonstrated that vaccination readiness—defined as personal attitudes toward COVID-19 vaccination—was the most important predictor of vaccination status. For each additional point on the Vaccination Readiness Scale, there is a 9.72 (95% CI 5.68-16.66; P<.001) times higher chance of being vaccinated. Another statistically significant influencing factor remains education; compared to people with a basic level of education, those with a medium level of education have a 6.78 (95% CI 1.76-26.15; P=.005) times higher chance, and people with a higher level of education have an 11.90 (95% CI 2.39-59.18; P=.002) times higher chance of being vaccinated. In addition, men have a 65% (95% CI 14%-86%; P=.02) lower chance of being vaccinated than women. Knowledge about vaccination, health literacy, external pressure, and self-efficacy did not play a significant role in the overall model.

In our study, we observed different information behavior regarding the COVID-19 vaccination, indicating not only the use and nonuse of such information but also the role of internet- and social media–based information. According to Baumann’s [8] model of information behavior, the groups were formed based on search frequency, motivation for searching, and channels used. The 3 user groups identified through cluster analyses are clear examples of cumulative strategies. The frequent and critical information evaluators (user group 1) rely on internet-based information of public authorities, whereas for the interaction-focused users (user group 3), social media–based information plays an important role. Furthermore, the infrequent and passive recipients reflect population groups (user group 2) who use information on COVID-19 vaccination less frequently. The observed types were similar to those described by Tan et al [12] and Piltch-Loeb et al [11], suggesting a degree of consistency in the results. Note that different variables were used in these studies to form the clusters and identify influencing factors. Frequent and critical information evaluators exhibited similar information behavior to the user groups “legacy” and “traditional omnivores” in Piltch-Loeb et al’s [11] study and “pro-formal selective” and “broad-trust” in Tan et al’s [12] study. Each of these user groups increasingly relied on local, traditional, and public channels from authorities to search for information, while making less use of social media and messaging services. In terms of vaccination behavior and attitudes, individuals in these groups were generally more willing to be vaccinated and demonstrated higher vaccination rates [11]. People with a strong intention to vaccinate and a positive attitude toward vaccination tend to use official websites more frequently, engage in more thorough information-seeking, and conduct searches more often [15,16].

There are parallels between the infrequent and passive recipients and the "nonseekers" in Piltch-Loeb et al’s [11] study or "broad distrust" in Tan et al’s [12] study. These groups rarely, if ever, actively searched for information and either did not use any channels or lacked trust in them. Consequently, these groups exhibited a low probability of vaccination and an ambivalent attitude toward COVID-19 vaccination.

The frequent and multichannel, interaction-focused users shared similarities with the "omnivore" + "broad social media users" and the "legacy and Facebook/Instagram users" [11]. Individuals in these user groups sought information about COVID-19 vaccination from both traditional public sources, such as government authorities, and increasingly from social media platforms such as Facebook and Instagram, as well as messaging services like WhatsApp [11]. Additionally, these user types tended to exhibit a higher likelihood of being unvaccinated and often held an undecided or negative attitude toward vaccination [11,12]. This trend may be linked to the observation that individuals who rely more on social media platforms for information rather than official websites generally have lower vaccination rates and a more negative view of vaccination [15,16]. The quality of information on these social media sources is often heterogeneous, and misinformation can spread rapidly, leading to uncertainty, skepticism, and rejection [32,33].

Frequent and critical information evaluators predominantly consisted of participants older than 44 years, who increasingly sought information about COVID-19 vaccination from traditional public and official sources [11]. In contrast, respondents aged 18-29 years were more likely to belong to the infrequent and passive recipients group. Individuals in this age bracket tend to engage in minimal or no active searching for information; when they do seek information, they adopt a targeted and specific approach [8,11].

The results of the analysis indicate that people with higher health literacy (user groups 1 and 3) searched for information on the internet more frequently than people with lower health literacy (user group 2). These are particularly groups who state that the reason for their search is to find out about current developments and to verify information. The study by Moffett et al [9] showed that people who are undecided about vaccination are more likely to conduct online research. This could not be confirmed in our study. However, a kind of uncertainty aspect may also have played a role among the frequent users identified here.

Notably, no clear correlation was identified between the user groups and vaccination status. Consequently, the determinants of information seeking identified by Baumann [8]—frequency, channels, and motivations underlying information-seeking activities—do not seem to be the primary determinants in this context. Instead, preliminary evidence suggests that the quality of the information accessed and the trust placed in it may influence vaccination behavior—variables that were not addressed in this investigation. Furthermore, only a few social differences could be observed, namely in gender and education. The vaccination readiness emerged as the strongest predictor of vaccination behavior in the analysis presented, highlighting the significance of individual beliefs and attitudes. These findings are consistent with previous studies that emphasized risk perception and social influence on vaccination behavior, which are components of the 7C Vaccination Readiness Scale used in this study [7,20].

Overall, when interpreting the results, it is important to consider that the study’s sample exhibited a higher vaccination rate of 86% (860/1000) compared to the 70% vaccination rate reported by the Robert Koch Institute for all citizens in Germany at the time of the survey [5]. Furthermore, the method of participant selection—the sample’s distribution—can also explain the discrepancies between the study’s findings and existing theories regarding the factors influencing information-seeking and vaccination behavior. The survey was conducted online by an external provider, which limited participation to individuals with internet access and registration with that provider. This limitation may impact representativeness, as those who are online are more likely to search for health information. Additionally, the health literacy score was based on subjective assessments, potentially leading to bias due to overestimation by the respondents. Moreover, no causal conclusions can be drawn from the cross-sectional design.

Several recommendations can be made for health communication regarding COVID-19 vaccination that may also apply to future novel infectious diseases. Identifying distinct types of online information use allows for a more nuanced understanding of information behavior. Targeted health communication to different user groups may not influence the vaccination behavior, but could lead to an informed vaccination decision, for example, understanding the risks and benefits. The following recommendations support a targeted approach for providing relevant health information to the identified user groups. For frequent and critical information evaluators, more evidence-based and detailed information could be provided on government and scientific platforms. Infrequent and passive recipients could be more effectively reached through low-threshold, easily understandable formats. In contrast, interactive formats, such as those found in social media or discussion forums, could better engage frequent and multichannel, interaction-focused users. Because this group obtains information from platforms that historically exhibit lower information accuracy, public health agencies should implement targeted interventions. For example, collaboration with trusted microinfluencers and platform‑native content creators could help coproduce short, evidence‑based messages that match the stylistic norms of each channel.

The 3 information-seeking groups identified in our study reflect fundamental behavioral patterns in how individuals obtain and process health information. Because these clusters were derived from usage‑based variables, they have the potential to serve as a generalizable framework for other health‑related issues. However, the transferability of the cluster structure is not automatic. For each new health topic, researchers should collect the same behavioral variables, conduct a confirmatory cluster or latent‑class analysis to test whether the original 3‑class solution fits the new data, and examine how cluster membership relates to the specific health behavior of interest. When the clusters are confirmed, targeted communication strategies that proved effective for COVID‑19 vaccination can be adapted to the new context.

Longitudinal studies are required to assess whether communication strategies that are tailored to each group produce sustained changes in knowledge, attitudes, and subsequent health actions. Finally, the efficacy of the proposed communication strategies should be evaluated through randomized controlled trials, allowing for evidence‑based optimization of future communication campaigns.