Having been identified as one of the most concerning global health threats according to the WHO (WHO, 2019), the WHO Vaccine Hesitancy Group describes vaccination hesitancy as the “delay in acceptance or refusal of vaccines despite availability of vaccine services” (WHO, 2014), which is subject to a number of factors, including “complacency, convenience, and confidence” (WHO, 2014). For 90% of countries worldwide, vaccination hesitancy poses challenges to national healthcare systems (Lane et al., 2018). In consequence, vaccination rates can be at best described as being “insufficient” (Habersaat and Jackson, 2020).

Classified as a people-made crisis by the WHO (2019), vaccination hesitancy is installed by a number of factors, including misinformation (Hussain et al., 2018), a lack of trust in and poor interaction with health professionals (Benninghoff et al., 2020), and movements towards more natural lifestyles, which reduce the necessity to vaccinate (Reich, 2016; Atwell et al., 2018). Vaccination hesitancy is even triggered by messages that strongly advocate vaccination uptake, which can induce highly skeptical individuals to reject vaccines even more strongly (Roose, 2020). The same was found to hold true for messages that contain a lot of information (WHO, 2019).

Vaccination-hesitant individuals have been found to have specific health information needs: they are best described as “active information seekers”, who search for “balanced information” (Andre et al., 2008; Greenwood, 2014). They are eager to arrive at informed decisions by looking for arguments both in favor of and against vaccination (Andre et al., 2008; Greenwood, 2014). Given their specific health information needs, they also consult information that is released by parties other than health authorities, and are particularly drawn to scientific studies to aid their decision making (Ball, 2020).

Apart from increasing incidents of anti-vaccine movements (Dube et al., 2020), the ongoing vaccination controversy has repeatedly addressed the risks and side effects associated with vaccines (Buts, 2020) or the skepticism towards scientific evidence regarding the effectiveness of vaccines (Larson et al., 2011; Badur et al., 2020). On the Internet, science critics (usually in the context of misinformation, fake news and denialism) coexist with science supporters (usually in the context of science advocacy and pro-science arguments) (Erviti et al., 2020; Agergaard et al., 2020). In recent years, individuals’ trust in the validity of scientific findings has decreased considerably (Browne et al., 2015), which is accompanied by a rise in non-scientific approaches (“denialism”) (Diethlem and McKee, 2009). This is conditioned by a “profound distrust in elites and experts” (Kennedy, 2019). Skeptical individuals doubt the validity of the information presented in advertisements or on websites and feel inclined to accept others’ opinions instead of forming their own (Sayed Hussin and Iskandar, 2015).

Health communication is often unable to produce its intended effects (Wilde, 1993; Foxcraft et al., 1997; Dillard and Shen, 2005), and might even lead to adverse effects (Hornik, 2002). Some of these undesired responses are grounded in individuals’ skepticism levels (Diehl et al., 2007; Koinig et al., 2018). Skepticism defines individuals‘ tendencies to meet information with disbelief (Obermiller and Spanberg, 1998), and is concerned with individuals’ willingness to believe (and consider) or disbelieve (and dismiss) message claims (Obermiller and Spanberg, 1998; Obermiller et al., 2005). In the health communication context, studies have determined skepticism to be negatively associated with drug involvement, comprehensibility of ad content as well as advertising as a source of health information in general (Diehl et al., 2007; Tan and Tan, 2007; Huh et al., 2012). This leads us to hypothesize:

H1: The higher individuals’ skepticism, the more negative are their message evaluations.

As individuals have become skeptical and distrusting of government officials, doctors and pharmaceutical companies (Schnirring, 2010), the validity of scientific findings is questioned more frequently (Kennedy, 2019). This phenomenon has been commonly referred to as reactance (Dilldard and Shen, 2005). Reactance thereby describes the failures e.g. in persuasive health communication to achieve its intended goals (Brehm and Brehm, 1981; Richards and Banas, 2015). While any persuasive message is able to motivate individuals to act upon the advocacy included therein, some messages might be met with reactance (Ringold, 2002), which occurs “when [individual] freedom is eliminated or threatened with elimination” (Brehm and Brehm, 1981). Reactance then originates out of individuals’ need for self-determination and autonomy (Burgoon et al., 2002). It can be triggered by a variety of message features, such as counter arguments or weak reasoning (Dillard and Shen, 2005). Some authors even go as far as claiming that any message that intends to change individuals’ behaviors can install reactance in individuals (Burgoon et al., 2002), who might fear that their freedom of choice is threatened. Based on the information presented above, we deduce the following hypothesis:

H2: Skepticism and Reactance are (positively) correlated.

Following the professional skepticism scale (Sayed Hussin and Iskandar, 2015), which is specifically used in the financial context, skepticism is made up of six different traits: 1) questioning mind (i.e. an openness towards the validity of the information presented); 2) suspension of judgment (i.e. evaluations are not made in a rush but based on a continuous assessment); 3) searching for knowledge (i.e. curiosity drives individuals to search for information, and ultimately, form attitudes), 4) understanding interpersonal relationships (i.e. the motivations of the source or message sender are questioned), 5) self-determination (i.e. the active choices made by the person when evaluating the content), and 6) self-confidence (i.e. the extent to which a person’s self-esteem drives their evaluations and judgments). The scale was chosen as a point of reference for the present study, since the items of the scale can be also applied to health-related information.

Research has confirmed that skepticism is indeed a powerful construct. For instance, in the advertising context, skepticism has induced individuals to change their responses towards advertising content (Obermiller and Spangenberg, 1998), and has been linked to various other constructs, such as “attitude toward the ad, believability of ad claims, perceived influence of the ads, and perceived untruths in the ads” (Obermiller and Spangenberg, 1998: p. 10). For this reason, we also expect individual responses to TBE information to be moderated by their skepticism levels (Obermiller and Spangenberg, 1998; Obermiller et al., 2005). As skepticism is negatively associated with comprehensibility of ad content as well as advertising as a health information source (Diehl et al., 2007; Huh et al., 2012), we expect highly skeptical individuals to exhibit higher levels of reactance, while less skeptical individuals are presumed to experience lower levels of reactance towards message content.

H3: Individuals’ message evaluations will influence their reactance to message content.

The goal of our study is to determine whether individuals’ responses to online information on TBE are influenced by their skepticism levels, which are characteristic of the present-day vaccine communication environment (Kahan, 2017). Thereby, individuals’ message evaluations will be used to determine their immediate message responses. Message evaluation is defined as individuals’ judgement of message characteristics and feelings as experienced during message reception (Burke and Edell, 1989). In a second step, individuals’ reactance towards message content will be determined, which describes their rejection of message content (Richards and Banas, 2015).

For this purpose, we conducted an online survey, in which individuals were exposed to information on TBE. The website utilized for the purpose of the study was a government-run website, which was in parts funded by a pharmaceutical company. The information presented on the website predominantly focused on the risks associated with tick bites. After reading the texts, participants were asked several questions regarding message comprehensibility and evaluation, as well as if the texts evoked some form of reactance in them.

We conducted an online survey and investigated whether there is a connection between the previously introduced constructs–skepticism and reactance. In that vein, we suppose, that individuals’ levels of skepticism influence their evaluations of selected text elements from the Austrian website on TBE. Skepticism has been found to influence a number of affective and cognitive variables (Kessler and Zillich, 2019), including message evaluation and reactance. Hence, in our view, skepticism constitutes a trait. Based on this assumption, we propose individuals’ levels of skepticism to influence their message evaluations and reactance levels respectively. Moreover, we propose skepticism and reactance to be correlated.

We carefully selected our text material for the online questionnaire, which was taken from the Austrian website on TBE (www.zecken.at). This website is hosted by a pharmaceutical company, but the overall initiative is a joint endeavor with the Austrian Federal Ministry of Social Affairs, Health Care and Consumer Protection. We used an eye-tracking study to identify those texts on the website that drew readers’ attention. We asked participants (n = 15) to gather information on TBE, giving them up to 10 min to search the website. Additionally, we controlled the experiment by asking other participants to gather information on bee-friendly gardening. Using both examples, we were able to analyze if there were significant differences between individuals’ eye movements, which was not the case. We used scan paths and heatmaps to identify relevant text elements. If an area (and text) did not draw any (zero) attention in terms of looking or reading (scan paths show reading patterns), the specific text was not considered for the survey. If participants were included in the pre-test of the study, they were excluded from participating in the main study.

The stimulus material itself was likely to trigger reactance in respondents, since it contained some controlling language. For instance, the text highlights the risks associated with tick bites—e.g. severe and long-lasting symptoms—and also lists a number of severe illness forms. Even though it does not explicitly threaten individuals’ freedom per se, the messages prompt individuals to get vaccinated to prevent the explicitly listed negative consequences of tick bites from occurring. By presenting several negative scenarios, and individuals’ vulnerability to tick-bites, the text induces a threat to individual freedom. We also included a link to the source, from where the text was retrieved.

The answers to each question were reported on a 7-point Likert scale ranging from (1) “I do not agree at all” to (7) “I fully agree.” Factor analyses revealed the items of the all multi-item variables to load on one single factor and to have acceptable Cronbach α values. Thus, they were combined for analysis:

Attitude Towards Vaccinations | Vaccination Hesitancy was measured with 5C psychological antecedents of vaccinations scale (Betsch et al., 2018). Vaccination confidence was measured with two items (based on Betsch’s confidence construct; KMO = 0.500, p = 0.000; α = 0.919), while vaccination hesitancy was determined via two additional items (based on Betsch’s complacency and constraints constructs; KMO = 0.500, p = 0.000; α = 0.729).

Skepticism was inquired by use of the Professional Skepticism scale (Sayed Hussin and Iskander, 2015) and consisted of 8 items (KMO = 0.746, p = 0.000, α = 0.729).

Message Evaluation was determined through 8 items (McKenzie and Lutz, 1989) (KMO = 0.751, p = 0.000, α = 0.782).

Reactance was based on the construct of psychological reactance and was measured by four questions (Dillard and Shen, 2005) (KMO = 0.760, p = 0.000; α = 0.856).

Additionally, we determined individuals’ preferred sources of (health) information to justify our focus on online health information. Sources of health Information were measured with single-item questions inquiring whether individuals used selected sources of health information. While most studies have used open questions, the present study followed previous examples (Stephens et al., 2004) by explicitly listing a number of sources which respondents could choose from.

Further we included a single-item (bi-polar) question to test respondents’ familiarity with the subject area, which we listed in full as well in abbreviated form: Are you familiar with the term Tick-Borne Encephalitis (TBE)?

Supplementary Appendix Table S1 provides an overview of the individual constructs and the corresponding items.

Subjects for the study were recruited via sending out links and using the snowball principle. This non-probability sampling method leads to a non-student convenience pool. We asked students from a quantitative method class, which was conducted at a medium-sized university in central Europe, to distribute the questionnaire through social media or email to generate a diverse pool of responses. While this sample does not allow us to draw conclusions for the overall Austrian population, it does, however, ensure a higher degree of heterogeneity than a sample that is solely based on students (Leiner, 2016). Further, as we seek to investigate whether individuals’ levels of skepticism impact respondents’ evaluations of the TBE information, we are still able to derive viable conclusions regarding potential differences among a more diverse sample.

After determining individuals’ levels of skepticism as well as their attitudes towards vaccination, the questionnaire ascertained respondents’ familiarity with the term TBE. Regardless of their answer, individuals were presented with a definition in order to ensure an equal state of knowledge before exposing them to the stimulus texts. After reading through the texts questions related to individuals’ message evaluations were posed, before inquiring individuals’ levels of reactance to the stimulus texts and the information presented therein. The questionnaire concluded with some demographic questions.

In total, 271 subjects were recruited to participate in the study. In terms of age, respondents were between 18 and 80 years old (M = 36.3 years, SD = 13.50). Approximately half of the sample was made up of students, while the other half of the sample was employed. More than 80% of the sample consisted of Austrian citizens. With regard to gender distribution, the largest part of the sample was made up of women (f = 65.7%; m = 34.3%), who are renowned to be more invested in health-related matters (Broom et al., 2009).

When inquiring individuals’ familiarity with TBE, the majority of subjects (80.8%) indicated that they had heard the term before. With regard to their preferred sources of health information, non-media sources (e.g., family or peer group members, doctors, or pharmacists) were more commonly consulted than media sources (e.g., print media, radio/TV or the Internet) (Niedereppe et al., 2007). Nevertheless, media sources remain of importance; this is also backed by more recent research, which found that the so-called patient information landscape is made up of both people (non-media) and media sources (Kantar Media, 2017). Out of the number of media channels provided, the highest scores were obtained for government-run websites (M = 4.01, SD = 1.88) and the Internet in general (M = 3.46, SD = 1.76), supporting previous findings (Betsch et al., 2010; Kessler and Zillich, 2019). While scores are below average (the scale’s mid-point), findings still support the notion that if media sources are used (as complementary sources of information) (Ruppel and Rains, 2012), people usually consult the Internet and Internet-based media sources.

In order to uncover respondents’ attitudes towards vaccination and confirm a potential vaccination hesitancy, two sets of questions were posed, one in favor of vaccination the other opposing vaccination. Overall, respondents seemed to advocate vaccination (M = 5.11; SD = 1.81) rather than rejecting them (M = 2.32, p = 1.45). Highly significant differences in respondents’ answers are noteworthy, whereby a clear and significantly higher tendency in favor of vaccination could be confirmed (T = 16.081, p = 0.000), which also reflects the German population’s mindset: 58% and 19% of respondents respectively were found to strongly or somewhat support vaccination (Statista, 2020b).

Hypothesis 1 postulated that highly skeptical individuals would have a more negative evaluation of the message dealing with the consequences of tick-borne encephalitis. Results indicate that, in general, individuals seem to be somewhat skeptical of online health information. Nonetheless, scores are slightly below the scale’s midpoint (M = 3.97, SD = 0.64), suggesting that the credibility and validity of online information is rarely doubted. This low score might be indicative of the fact that—particularly in the context of vaccine-related health decisions—individuals feel rather confident to judge the accuracy of online health information.

When testing whether individual skepticism had an effect on individuals’ message evaluations, results of a linear regression fail to account for the presumed effect. The model fit turned out to be not significant (R ² = −0.001, F = 0.741, p = 0.390). In this case, we have to consider that skepticism is a rather stable predisposition as one’s skepticism does only marginally influence the direct perception of messages. Nonetheless, hypothesis 1 is confirmed.

According to hypothesis 2, individuals’ skepticism and reactance would be positively correlated. This would mean that if individuals were skeptical of scientific or health information, they would be more inclined to question the information presented on TBE websites. Surprisingly, results do not support this assumption, and instead of a presumed relationship, we were not able to detect any correlation (r = −0.022, p = 0.747). This result might be explained as follows: since reactance usually triggers emotional reactions, the informative arguments presented on the website might surprise individuals, but not appeal to them emotionally. In consequence, they do not feel deprived of their control, and thus, are not likely to enter into a state of reactance (Dillard and Shen, 2005). Consequently, hypothesis 2 is rejected.

Hypothesis 3 presumed that message evaluations would influence individuals’ reactance to message content. We predict responses to be subject to individuals’ skepticism. Hypothesis 3a suggested highly skeptical individuals to hold more negative message evaluations, which would lead them to oppose the message arguments rather strongly and, thus, lead them to exhibit higher levels of reactance. In line with hypothesis 3a, hypothesis 3b presumed the opposite effect for less skeptical individuals: their message evaluations are expected to be more favorable, and therefore, we presume them to exhibit lower levels of reactance to message content.

We conducted a regression analysis and added age and gender as controlling variables. In this case the model fit turned out to be significant (F (3,267) = 7.263, p = 0.000). Neither gender (t = 0.013, p = 0.999) nor age (t = 0.766, p = 0.444) influenced reactance, yet we found a negative message evaluation (t = −4.624, p = 0.000) to result in a higher level of reactance. Nevertheless, the R ² = 0.065 is not significant at all, suggesting that these three variables are not able to determine the level of reactance sufficiently. When splitting the data into two groups (low skepticism vs. high skepticism), we found that the effect to be more pronounced for the less skeptical group (R ² = 0.192; F = 8.384, p = 0.000) than for the more skeptical group (R ² = 0.096; F = 3.734, p = 0.013). We are thus able to confirm both hypotheses 3a and 3b.