**3. Results**

In order to determine the influence of the analyzed factors on parents' attitudes toward children's vaccination and their confidence in vaccines, an analysis of the characteristics of the respondents was performed using the descriptive statistics method, which is presented in Table 1.


**Table 1.** Characteristics of respondents.

Based on the descriptive statistics provided in Table 1, it may be concluded that the respondents are predominantly female (as much as 70.1% of the total number of respondents, which may indicate a greater interest of mothers in the mentioned topic), with the largest number between 35 and 40 years (40.7% of respondents). Respondents are married in 89.7% and possess a faculty diploma in 40.1%. The largest number of respondents is from Montenegro (44.3% of respondents, which may be justified by the fact that the research was initiated in this country) and have two children (45.1%). In terms of geographical spread, the demographic of the respondents is as follows: 705 (44.30%) respondents are from Montenegro and 520 (32.60%) from Serbia, while 368 (23.10%) are from Bosnia and Herzegovina.

Furthermore, these characteristics were correlated with the degree of trust that parents have in vaccines, which is shown in Table 2.

**Table 2.** Table of contingency for attitudes, i.e., trust in vaccines in relation to the characteristics of the respondents.


The greatest dispersion of data about the average value of the attitude that they do not trust in vaccines was noticed in the question related to the age of the respondents. The standard deviation of this characteristic is 1359. Thus, the age of a randomly selected sample deviates from the average value of all respondents in the amount of 1359 points on the Likert scale, provided that these subjects do not trust vaccines. High data dispersion was also noted for the level of education of respondents in this category of trust in vaccines because the value of the standard deviation is 1006. The two highest values in category I, totally believe, are for education (1473) and age (1191). The analysis of other categories of trust in vaccines also showed that the results are the most dispersed for the level of education and for age because standard deviations have the highest values for these two characteristics of the respondents. The most homogeneous answers refer to the gender and marital status of the respondents.

Based on the contingency table, a graphical presentation (Figure 2) of the participation of individual categories of respondents in terms of characteristics, such as gender, country of origin, age, marital status, and number of children, in combination with their attitude about trust in vaccines was created.

**Figure 2.** Trust in vaccines.

The analysis of the contingency table showed that the largest number of respondents have high confidence in vaccines (34%), when their characteristics, such as gender, country of origin, age, marital status, and number of children, would be analyzed. However, about one-fifth of respondents do not trust or have a low level of trust in vaccines, which means that certain actions must be taken in order to improve this situation. This was the motive to focus on discovering the reasons why parents do not trust vaccines in one part of the research in order to create set of recommendations which would contribute to the strengthening of trust in vaccines.

Furthermore, characteristics of respondents have been correlated with the attitudes of parents towards vaccines, as shown in Table 3. Since the survey was conducted in three countries, three groups of respondents were available for testing, so it was possible to apply the ANOVA test.


**Table 3.** ANOVA test of determination of differences based on characteristics of respondents.

Symbol \* represents the combination of two variables.

The starting hypothesis of the ANOVA test indicates the equality of expected values for the characteristics of the respondents, such as gender, age, country of origin, level of education, marital status, and number of children, which the respondents have in relation to their confidence in vaccines. The analysis of the variance of the respondents' data on the above characteristics, given in the previous table, has shown that the expected value for each individual characteristic (except for the state) differs in relation to the attitudes of parents, and that the given characteristics have a significant influence on the formation of attitudes. Based on the obtained results, it is possible to accept hypothesis H1.

Using the conclusion of hypothesis H1, and before testing the justification of the claim of hypothesis H2, it was tried to answer the question of the relationship between the demographic characteristics of respondents (who were the subject of hypothesis H1) and their negative attitude towards vaccines based on information from online media (correlated with hypothesis H2) and for that purpose logistic regression was applied. So, in the continuation of the research, we analyzed the influence of certain characteristics of parents on their attitude not to vaccinate a child, built on the content, which they found by consulting online media. The aim of this part of the analysis was to determine whether there is a certain group of respondents who are more vulnerable to content on online media and, on that basis, refuse to vaccinate a child. In order to determine the relationship between a particular characteristic of respondents and their attitude not to vaccinate a child, under the influence of information found on online media, as mentioned above, we used logistic regression because it is most often used to rank the relative importance of independent variables and to quantify the effect their interactions.

In order to define the variable which represents the negative attitude of parents towards vaccines, formed on the basis of content from online media, we chose the answer to one key question in the survey, which represents this behavior. To define the anti-vaccine attitude, we considered the answer to the question "Texts on online media about the negative effects of the vaccine affect the formation of my attitude to a significant extent." because we believe that other questions about the negative attitude towards vaccines formed on the basis of online media are less focused on forming an attitude and making the final decision not to vaccinate the child. The independent variables in the model are the following key characteristics of the respondents: gender, age, country of origin, level of education, marital status, and number of children in the family. Thus, by assessing logistic regressions, we tried to find an answer to the question of whether men or women, younger or older parents, parents

with higher or lower education, etc., are more prone to negative attitudes towards vaccines based on information obtained from online media. The results of the analysis are given below.

Before analyzing the model, we examined its quality by testing the hypothesis that there is no relationship between the dependent and independent variables in logistic regression. The test results are given in the following table (Table 4).

**Table 4.** Model fitting test for logistic regression of negative attitude of parents towards vaccines formed on the basis of information from online media.


In this case, we tested the model by comparing the initial value of the logarithm, i.e., the model without an independent variable, which is 2896.054 with the final model, i.e., the model with an independent variable, which is 2401.083. With 72 degrees of freedom, χ2 is 494,971, which is significant at the level of 0%. The obtained results show that the model is meaningful and that the null hypothesis about the non-existence of a connection between the independent and dependent variables cannot be accepted.

The results of the evaluation of the logistic regression model are given in the following table (Table 5).



At the very beginning, it should be emphasized that all parameters in the regression are statistically significant with a risk of error of 5%. Since we are most interested in commenting on the results of extreme values on the Likert scale, we defined the value 1 as a basis for comparison (I completely disagree), and the value to explain the relationship between demographic characteristics and anti-vaccine attitude based on information from online media was defined with value 5 on the Likert scale (I completely agree). Based on the results of the estimated logistic regression, it is concluded that, if the parent-respondent is a male, i.e., the father, in 69.7% of cases, he will less often form a negative attitude towards vaccines based on information or texts read on online media compared to mothers-respondents. In other words, negative texts about vaccines through online media are not key to forming a negative attitude among fathers-respondents. If the age of the respondents' parents is observed, conclusions on this issue are made on the basis of the reference group of parents, who are older than 45 years. If parents are under the age of 35, they are more likely to form a negative attitude towards vaccines and eventually make the decision not to vaccinate their child by reading texts through online media compared to parents over the age of 45. On the other hand, parents who are between 35 and 45 years old are about 80% less likely to form a negative attitude towards vaccines based on information from online media compared to the parents of the oldest age group of respondents in this survey. Parents from Serbia and Montenegro have between 15 and 18% less chance of forming a negative attitude towards vaccines reading articles on online media compared to parents from Bosnia and Herzegovina. Parents who have any lower level of education than doctors of science, are more likely to form a negative attitude towards vaccines based on information from the online media compared to parents with the title of doctor of science. This conclusion makes sense because parents with the highest level of education are more inclined to check the information and thoroughly process each topic before making such an important decision, such as vaccinating children. Parents who are married, in common law marriage, or divorced are more likely to build a negative attitude towards vaccines by reading content on online media compared to parents who are widowed. And that chance is incomparably higher for parents who are in common law marriage compared to other marital statuses. Finally, if the number of children in a family is observed, parents with less than three children are between 50% and 80% less likely to form a negative attitude towards vaccines based on texts they found on the Internet than parents with three or more children. This conclusion is connected with the fact that parents with a smaller number of children will strive to obtain additional information about vaccines from other sources, so the position on vaccines will not be formed only on the basis of content from the Internet, while parents with more children have a di fferent situation.

Based on this part of the research it may be concluded that the following groups of parents are particularly vulnerable to the influence of online media on attitudes toward vaccines: women, parents of younger age ("millennials"), and parents who are in common law marriage, as well as parents who have more children.

Furthermore, in order to determine the level of influence that online media have on parents' attitudes, an ANOVA test was applied to test the second hypothesis. The results are given in Table 6.



An analysis of the equality of expectancies for the characteristics of research on parents' attitudes toward vaccines in the situation when using online media as a source of information has shown that respondents' attitudes differ significantly depending on which aspect is used as a source of information. The conclusion is that online media has a significant influence on the formation of parents' attitudes toward the vaccination of children, which leads to the acceptance of the hypothesis H2. In addition, the logistic regression itself showed that there is a significant difference in the formation of a negative attitude towards vaccines in different demographic groups of respondents, which is formed on the basis of information from online media.

In order to further examine the importance of a particular way of using online media on formation of attitudes, an eta coefficient was used, in which the squared value represents a relative measure of association. The eta coefficient takes a value between 0 and 1 and represents the proportion of variance in the dependent variable explained by the independent variable. The formula for calculating the eta coefficient is:

$$
\eta^2 = \frac{\text{Safety}}{\text{SStotal}}
$$

where:

> SSeffect = the sum of squares for a given independent variable (factor);

SStotal = the total sum of squares for all factors, interactions, and errors in the ANOVA analysis. If η2 is 0.01, then it indicates a small impact, while a moderate impact is indicated for a value of 0.06 and a large influence for a value of 0.14 or greater. The results of the eta coefficient are given in Table 7.


**Table 7.** Measures of association of online media with attitudes toward vaccines.

The analysis of the η2 coefficient has shown that the greatest value in the opinion of parents that the texts on the online media about the negative effects of the vaccine influence the formation of the parents' attitude (η<sup>2</sup> = 0.216). Then, there is the opinion that these texts are generally correct (η<sup>2</sup> = 0.181), and thirdly, is the view of parents that these texts have no scientific basis (η<sup>2</sup> = 0.167), that is, they are based on fears and speculations. On the other hand, when claiming that the texts on the online media about negative effects have more importance than they should have, the value of the η2 coefficient is significantly 0.090.

In order to examine whether there are differences in the analyzed countries regarding the influence of online media on the formation of attitudes, additional testing has been done and the results are shown in Table 8.


**Table 8.** ANOVA test examinations of di fferences by analyzed countries.

Symbol \* represents the combination of two variables.

The analysis of variance showed that the attitudes of the interviewed parents were the same, regardless of whether the respondents were from Montenegro, Serbia, or Bosnia and Herzegovina. Namely, the error made by rejecting the hypothesis that parents' attitude toward trust in vaccines is equal to 98.7%, which is why we cannot reject it. In addition, in the section that follows (Table 9), an analysis of the eta squares for states in relation to attitudes, i.e., trust in vaccines, is presented.

**Table 9.** Measures of association of countries with attitudes of respondents.


The measure of association, that is, the eta coe fficient for the country from which respondents come in relation to respondents' attitudes toward vaccination is 0. In other words, there is no statistically significant di fference between the analyzed countries for explaining the influence of online media on parents' attitudes toward vaccines. Based on the results obtained previously, it may be concluded that hypothesis H3 is confirmed. Hence, the ANOVA test and the eta coe fficient confirmed that the country of origin of the respondents does not a ffect their attitude towards the vaccination of children.
