A cross-sectional web-based survey was conducted from November 1 to November 5, 2021, among people aged 20-69 years. The participants were recruited online using Surveroid (Marketing Applications Inc). The primary analysis was a correlation analysis for each age group separated by 10 years. We used the Cohen medium effect size because we could not find any previous reference [23]. Thus, the effect size was set at r=0.3, with an alpha set at .05 and a beta set at .2. The required sample size was 84 participants per age group. Surveroid collected an extra 20% of the sample in case of response errors or omissions. Therefore, the final number of participants was calculated to be 102-103 per age group, for a total sample size of 512. Equal numbers of men and women were assigned to each age group. The study was described at the beginning of the web-based questionnaire, and completion of the questionnaire was considered as consent to participate in the study. The web-based questionnaires were collected in a randomized identification format without asking for personal information such as names or email addresses. The participants received a reward upon completion based on their registration status on Surveroid database. A survey request was sent to 8809 people via email. The survey was terminated when the target number of 512 participants was reached.

The web-based questionnaire consisted of the following five factors or groups of items: (1) sociodemographic indicators and experiences during the COVID-19 pandemic, (2) COVID-19 prevention behaviors, (3) health literacy, (4) COVID-19 knowledge, and (5) fear of COVID-19. The questionnaire could be completed in 5 minutes.

The distribution of the studied variables was explored using descriptive analysis. The Kruskal-Wallis rank sum test was used to compare the scores of each item for each age group. If there was a significant difference in the Kruskal-Wallis rank sum test, the Tukey-Kramer test was added to determine which group had a significant difference. The relationship between BS, CCHL, FCV-19S, and CAS was reported using Spearman rank correlation analysis, as the assumptions of normality and homoscedasticity were violated for all indicators. In addition to the overall analysis, these analyses were performed separately for each age group to elucidate the differences. Next, multiple regression analysis was conducted for each age group, with BS as the dependent variable; CCHL, FCV-19S, and CAS as independent variables; and sex and age, which have been associated with health-related behaviors in previous studies, as adjustment variable [7,28]. When there was a correlation between the independent variables, multicollinearity (variance inflation factor [VIF]) was examined; if the VIF was <10, it was determined that there was no multicollinearity. The significance level for rejection of the null hypothesis was 5%. Statistical analysis was performed using the JMP Pro version 15.0 statistical software (JMP Statistical Discovery LLC).

The Medical Ethics Committee of Kyoto University, Japan, approved this study (#R3215).

A total of 512 adults aged ≥20 years participated in the survey. The mean age was 44.3 (SD 14.3) years. All age groups were assigned approximately equal numbers across men and women. The distributions of educational level, COVID-19 infection or contact experience, and number of vaccinations are shown in Table 3.

The results of the comparison of scores for each variable by age group are summarized in Table 4. The age group of 60-69 years had significantly higher BS than in 20-29 years. The age group of 60-69 years had a significantly higher CAS compared with the other age groups.

The results in Table 5 show that, for all participants, BS had significant positive correlations with CCHL, CAS, and FCV-19S; significant negative correlations with FCV-19S and CAS; and significant positive correlations with CCHL and CAS.

Analysis by age group shows that, in the 20-29 age group, there are significant positive correlations between BS, CCHL, and FCV-19S, and significant positive correlations between CCHL and CAS; in the 30-39 age group, there are significant positive correlations between BS, CCHL, and FCV-19S, and significant negative correlations between FCV-19S and CAS; in the 40-49 age group, there are significant negative correlations between BS and CCHL, and significant positive correlations between CAS and FCV-19S, and between FCV-19S and CCHL; in the 50-59 age group, BS and CAS have significant positive correlations, and CCHL and CAS have significant positive correlations; in the 60-69 age group, only CCHL and CAS have significant positive correlations.

The results of the multiple regression analysis with BS as the dependent variable and sex and age as the adjustment variable are presented in Table 6. In all cases, a VIF of <10 was found. Thus, we determined that there was no influence on the multiple regression analysis. In the analysis of all age groups, all variables were significant, and BS was associated with CCHL, FCV-19S, and CAS. By age group, the following variables were significantly related to BS: CCHL and FCV-19S at 20-29 years; CCHL and FCV-19S at 30-39 years; CCHL, CAS, and FCV-19S at 40-49 years; and FCV-19S at 50-59 and 60-69 years.

This study aimed to examine the association between COVID-19 prevention behaviors and health literacy, COVID-19 knowledge, and fear of COVID-19 by age group. Overall, CCHL, CAS, and FCV-19S significantly influenced COVID-19 prevention behaviors. We also found a significant negative correlation between FCV-19S and CAS, and a significant positive correlation between CCHL and CAS. The correlations between FCV-19S and CAS, as well as between CCHL and CAS were determined to have little impact on the multiple regression analysis as the VIF was examined. However, multiple regression analysis by age group showed that the factors associated with COVID-19 prevention behaviors differed by age group. Among the 20-29, 30-39, and 40-49 years, multiple factors, including health literacy, influenced COVID-19 prevention behaviors. However, at 50-59 and 60-69 years, FCV-19S influenced COVID-19 prevention behaviors, whereas CCHL and CAS did not.

Regarding the characteristics of the participants in this study, the educational level in all age groups was higher than in the census results published in 2015 [29]. Therefore, the participants were considered to have a higher level of education than the average Japanese population. Regarding the experience of infection or close contact with patients with COVID-19, most participants across all age groups answered “neither.” The published infection rate of COVID-19 in Japan on January 1, 2022, was 1.4% [15], which was almost the same as the infection rate in this study. Most participants (76.56%) had received both vaccine doses. The difference in the number of vaccinations according to age is thought to be due to priority vaccination for older adults and people with chronic diseases [22]. In addition, the mean CCHL score of 3.55 for all participants is comparable to that of previous studies [25,26] and is at the same level of health literacy as the average Japanese. In this study, CCHL was lower than average for those aged 20-29 and 30-39 years, which is consistent with previous studies showing that health literacy is higher in older age groups [21].

In this paper, we first discuss the overall trends. The results of this study that health literacy, knowledge, and fear influence prevention behavior were similar to the previous studies. Health literacy was assessed using the CCHL, which was created to assess the public health literacy of the average citizen, and consists of higher-order communicative and critical health literacy rather than functional health literacy [25]. Communicative literacy is the ability to extract and understand information from a variety of communications and actively apply new information to changing circumstances in everyday activities [30]. Critical literacy refers to the ability to critically analyze and use information to better control a situation [30]. A previous study has shown that the higher the communicative or critical health literacy, the healthier the lifestyle [25]. This finding suggests that a high level of communicative or critical health literacy enables people to obtain information from various sources. Furthermore, it is likely that they can analyze the information obtained, leading to better behaviors. Primary studies have shown a relationship between health literacy and COVID-19 knowledge [13], and this study also showed a positive correlation between health literacy and COVID-19 knowledge. Therefore, even in the COVID-19 infodemic, people with high health literacy were able to obtain true information and adopt COVID-19 prevention behaviors without being misled by misinformation. These findings suggest that improving health literacy could help promote COVID-19 prevention behaviors. In addition, it is important to create an information dissemination system that allows people with low health literacy to obtain accurate information. Previous studies have shown that Japanese people have lower health literacy than Europeans [21]. This is because it is difficult to search for evidence-based information in Japan, as searches for diseases and symptoms often show information that is not available on official websites [21]. We believe that tools that enable anyone to access true information easily are needed.

Fear of COVID-19 was also associated with COVID-19 prevention behaviors. A previous study showed that the more threatening COVID-19 is perceived to be, the more prevention behaviors are taken [24]. In this study, as in the previous study, the greater the fear of COVID-19, the more prevention behaviors were adopted.

As mentioned above, health literacy, COVID-19 knowledge, and fear of COVID-19 are all relevant to prevention behaviors. However, panic and fear due to COVID-19 have been associated with depression and high stress states [31]. The results revealed a negative correlation between COVID-19 knowledge and fear of COVID-19. Therefore, promoting prevention behavior through correct knowledge of COVID-19, rather than fear of COVID-19, is considered more advisable from a citizen's mental health perspective. Communicating correct information, rather than fearmongering, is considered more important.

Next, we discuss the results according to age group. Among the 20-29 years, 30-39 years, and 40-49 years age groups, multiple factors, including health literacy, influenced COVID-19 prevention behavior. By contrast, in the 50-59– and 60-69–years age groups, only FCV-19S influenced COVID-19 prevention behaviors, whereas CCHL and CAS did not. One reason might be that the older age group had experienced influenza pandemics in 1957 and 1968 as well as the emerging infectious disease, severe acute respiratory syndrome pandemic in 2002 [32,33], and thus more easily adopted prevention behaviors, regardless of their health literacy and COVID-19 knowledge. In addition, in the COVID-19 pandemic, there is another possibility—COVID-19 has been reported to cause more severe symptoms at older ages [15] and is routinely reported on television and in newspapers in Japan. Therefore, older individuals are more likely to perceive the risk of infection as a threat. Furthermore, experts argue that older adults are vulnerable to anxiety and stress caused by the pandemic and are particularly fearful of social isolation and loneliness [34,35]. Against this background, fear of COVID-19 may have been a factor in prevention behaviors in the older age group, regardless of health literacy or COVID-19 knowledge. However, the relationship between fear and prevention behaviors is a vicious cycle that may lead to isolation due to prevention behavior that avoids seclusion and worsens mental health due to fear. We believe that prevention behaviors need to be supported by correct knowledge and health literacy and not solely based on fear. To this end, social support to reduce the stress caused by the pandemic, to protect against social isolation, and to provide correct knowledge is considered important.

Mature adults, similar to the overall trend, had multiple factors associated with prevention behaviors, including health literacy. One of the differences between younger and older adults is that social media use is higher among the younger age groups. Our contemporaneous study found that social media use was higher among those in their 20s and 30s than among the 50s and older age groups [36]. During the COVID-19 infodemics, high health literacy and correct knowledge may have been more important because a lot of misinformation was disseminated on SNS. Based on the results of this study, we believe that support tailored to age-specific characteristics is needed to promote prevention behaviors.

This study had several limitations. First, this was a cross-sectional study; therefore, causal relationships could not be established. Longitudinal studies must be conducted to confirm the causality. Second, the participants had a higher level of education than the general Japanese population. In addition, as this was a web-based survey, the participants were limited to internet users. Questionnaires should be administered to a wide range of individuals. Third, this study was conducted in November 2021, a period during which the number of COVID-19 cases in Japan had declined. The FCV-19S isolates in this study differed from those reported in previous studies [16,27]. The fear of COVID-19 may vary depending on the infection status and the type of virus prevalent. In addition, COVID-19 prevention behaviors may be altered by government policies. Therefore, further investigations of different infection statuses would be helpful. Fourth, this study collected questionnaires from participants only once. Therefore, the responses' reliability in this sample has not been verified. Therefore, we will need to examine the reliability of the responses in the future.

The COVID-19 pandemic is still ongoing, and we need to consider how to live with this virus. Governments and local authorities need to encourage people to take action to prevent the spread of infection and further lockdowns. The results of this study indicate that factors associated with prevention behavior differ by age, suggesting that there may be different means of promoting prevention behavior. However, information on COVID-19 is updated daily, making it difficult to generalize the results of this study. Therefore, it is necessary to expand the time period and geographical regions in future studies. In addition, longitudinal studies are needed to show causal relationships and to improve the level of evidence by providing health literacy education and interventions to alleviate fear.

In this study, we identified the factors that influence COVID-19 prevention behaviors and found that these factors differed by age. In the age groups 20-29 years, 30-39 years, and 40-49 years, multiple factors, including health literacy, influenced COVID-19 prevention behaviors, whereas in the age groups of 50-59 and 60-69 years, only fear of COVID-19 influenced behavior. This suggests that tailored support according to age and risk factors is needed to promote prevention behaviors.