The health behavior of college students is a topic that is worth exploring. According to Taiwan’s Health Promotion Administration of the Ministry of Health and Welfare, National Health Survey statistics showed that the percentage of college students aged 18-24 years who exercised regularly was 60.8% in 2002, 60.6% in 2005, and 55.1% in 2009. Furthermore, the percentage of students who ate breakfast daily was 65.8% in 2002, 62.6% in 2005, and 57.3% in 2009 [1]. Data from a 2010 survey showed a 2-hour difference between the number of hours students spent sleeping on school days (6.4 hours) and weekends (8.5 hours). These results indicate that Taiwanese college students’ exercise, eating, and sleep behaviors must be improved [2]. Many of the lifelong habits that jeopardize health are formed during childhood and adolescence [3]. As college students transition from adolescence into adulthood, their health habits may affect their future well-being. During this transition, those who have poor health habits may adopt better habits if they are given sound advice. Therefore, it is necessary and important to examine college students’ health behaviors because these behaviors affect students’ physical health and lifestyles in adulthood [4]. Health behavior is “any activity undertaken by a person who believes himself to be healthy, for the purpose of preventing disease or detecting it in an asymptomatic stage” [5]. A number of studies on health behavior have focused on eating and exercise [4,6]. Given that sleep is a basic physiological need and an essential element in maintaining physical and mental health, we will examine the health behaviors of college students by measuring their eating, exercise, and sleep behaviors.

Individual factors and eHealth literacy may affect one’s health behaviors. According to social cognitive theory [7], each factor possesses a self-regulating system that affects motivation and learner differentiation. Human behavior is influenced and affected by the individual, the behaviors of others, and the environment. This self-regulating system represents a process that is affected by bidirectional and interdependent associations between and among behaviors, environments, and personal experiences. Studies have found that certain factors, such as one’s health status, concern for health, and eHealth literacy, may shape an individual’s health behaviors [8].

eHealth literacy may mediate the association between demographic factors and health behaviors. The integrative model of eHealth use (IMeHU) suggests that the underlying social structure affects an individual’s health status, computer literacy, intrinsic interest in health, and perceived ability to use the Internet for health purposes [9]. The model also proposes macrolevel disparities in the social structures that are connected to health disparities through the microlevel conduits of eHealth literacy, motivation, and ability. That is, individuals with low eHealth literacy have less incentive to use the Internet to access health information and consider themselves to be incapable of using Internet-based health information. The IMeHU provides a parsimonious account of the connections among the digital divide, health care disparities, and the unequal distribution and use of communication technologies. However, few studies have examined the associations among individual factors, eHealth literacy, and health behaviors. Similarly, there is a lack of empirical evidence regarding the IMeHU. This study uses the IMeHU as a framework for examining the associations among individual factors, eHealth literacy, and health behaviors and to further validate the mediating effects of eHealth literacy on health behaviors and individual factors.

According to the Ministry of Education and Statistics Department in Taiwan [24], there were a total of 1,037,632 college students in 2012. Accordingly, a sample of 350 was suggested for this investigation [25]. We determined that 500 students were needed for the official sample and estimated an effective questionnaire return rate of 80%. To make the sample more representative, the stratified cluster sampling method was employed. Using the region on a tiered basis and the class as the sampling unit, we extracted the desired projected number for each region sample based on the proportion of university students in the northern (n=292), central (n=146), southern (n=176), and eastern regions (n=11) of Taiwan. From March to May 2013, we surveyed a nationally representative sample of college students. A total of 625 questionnaires were distributed, including 64 unreturned and 36 incomplete questionnaires. Thus, 525 usable questionnaires were collected, resulting in an effective response rate of 84%. The sample’s sociodemographic and health information is displayed in Table 1.

Finally, we gathered the respondents’ background information, including information about their health status, their major, their degree of health concern, and the frequency with which they engaged in health-related discussions.

The students’ health status level (perceived psychological and physical status) was measured by asking them how well they currently felt with respect to their own psychological and physical condition on a scale from 1 (strongly unwell) to 10 (strongly well).

The students’ degree of health concern was measured by asking them about their health concerns and willingness to take appropriate action. The degree of importance of participants’ health was rated on a scale from 1 (extremely unimportant) to 5 (extremely important). A higher score indicated a higher level of concern regarding one’s health.

The major dimension was divided based on whether the participants were majoring in medical fields. In a subsequent analysis, the 2 groups were then transformed into dummy variables. We used the nonmedical group as the reference group.

The frequency with which students engaged in health-related discussions was measured based on participants’ responses on a scale from 1 (seldom) to 5 (always).

This study used Amos 6.0 to perform a confirmatory factor analysis to identify the best measurement models for the eHLS and HBS. We then used multiple regression analysis to examine the associations among individual factors, eHealth literacy, and health behaviors. Additionally, a causal steps approach was used to explore the indirect (mediating) effects [36] and Sobel tests [37] were used to test the significance of the mediating effects.

Table 2 indicates that nearly all the individual factors positively predicted the 3 dimensions of eHealth literacy, yielding low predictive explanatory powers of 7%-8%. Notably, all the individual factors emerged as significant indicators of the 3 dimensions of eHealth literacy, with the exception of the frequency of health-related discussions.

Two multiple regression analyses were performed to examine how well individual factors and health literacy predicted health behaviors. Table 3 reveals that the individual factors positively predicted the 3 health behavior dimensions, with a moderate level of predictive explanatory power for eating behaviors (23%) and a low level of predictive explanatory power for exercise (13%) and sleep behaviors (13%).

Table 3 further reveals that both health status and health concern positively predicted the 3 health behavior dimensions; major and the frequency of health-related discussions only demonstrated positive predictive power for eating behaviors. In addition, functional and critical eHealth literacy displayed positive predictive power for eating and exercise behaviors, whereas critical eHealth literacy only positively predicted sleep behaviors.

Table 4 is the result of a regression analysis of the individual factors and eHealth literacy as predictors of health behaviors. When comparing Tables 2 and 3, the standardized regression coefficients of the independent variables either decreased or were insignificant. Furthermore, a Sobel test of the mediating effect indicated that 6 of 13 paths were significant.

The 2 standardized regression coefficients were multiplied to compute a mediating effect, as presented in Table 5. Critical eHealth literacy mediated the association between health status and eating, exercise and sleep behaviors, showing mediating effect values of .015, .021, and .022, respectively. Functional eHealth literacy mediated the association between health concern and eating and exercise behaviors, with mediating effect values of .022 and .021, respectively. Critical eHealth literacy mediated the association between health concern and eating behaviors, yielding a mediating effect value of .017.

According to Baron and Kenny’s approach to statistical mediation analysis [36], it was further determined that critical eHealth literacy mediated the association between health status and health behaviors (see Table 6) and that the mediating effect was .058 units (.015+.021+.022). Functional/critical eHealth literacy mediated the association between health concern and health behaviors, with a mediating effect of .060 units (.022+.021+.017). These results indicate that participants who had better health status and greater concern for their health tended to have better functional and critical health literacy and were, therefore, more inclined to engage in positive health behaviors, especially positive eating behaviors.

This study found that the participants who had better self-perceptions of their health status and stronger concern for their health exhibited better eHealth literacy and had an increased likelihood of adopting healthy eating, exercise, and sleep behaviors. Moreover, eHealth literacy mediated the association between individual factors and health behaviors.

This study found that the participants who had better perceived health status and greater concern for their health tended to adopt more positive health behaviors. Similarly, college students who majored in medical fields and who engaged in more health-related discussions demonstrated better eating behaviors. Thus, hypothesis 1 was largely supported. Consistent with social cognitive theory [7], an individual’s health status and health concern are key factors that prompt him or her to adopt health behaviors [10,12]. It appears that college students who have better perceived health status and greater concern for their own health pay greater attention to their health and are more willing to engage in appropriate health behaviors.

The study also revealed that the participants who majored in medical fields, had better perceived health status, and had greater concern for their own health tended to have better eHealth literacy than other students. Additionally, the participants who frequently engaged in health-related discussions had better critical eHealth literacy. Therefore, consistent with previous studies [22,23,38], hypothesis 3 was largely supported: college students who had better perceived health and paid more attention to their health were more likely to seek and evaluate health information.

The framework for health literacy identifies 3 major areas of potential intervention and forms the organizational principles of intervention [8]. The framework also illustrates that educational, health, and cultural and social factors may influence health literacy and may ultimately contribute to health outcomes and costs. Health literacy involves a range of social and individual factors, including cultural and conceptual knowledge, listening, speaking, writing, and reading skills. For example, a previous study found that performance on health literacy tasks was related to education, income, country of birth, age, and race/ethnicity [39]. Specifically, individuals with higher educational attainment and higher income demonstrate higher levels of health literacy [9]. For example, medical school students have higher incomes and more medical knowledge and, therefore, possess greater eHealth literacy than non-medical school students.

However, this study also found that the participant’s major and the frequency with which he/she engaged in health-related discussions did not predict exercise and sleep behaviors. Because eHealth literacy relates to one’s context of relevant medical knowledge when assessing the quality of health information and when making decisions that promote one’s health [9], students who major in medical fields are likely to have greater knowledge of medicine. Therefore, medical school students have better health literacy than non-medical school students. The knowledge, attitude, and practice (KAP) model advocates that once an individual receives relevant information, the individual will develop the expected responses, thus triggering behavior that is consistent with one’s attitude [40]. However, outcomes and efficacy are not necessarily aligned [7]. In other words, what one knows does not necessarily equate to what one does.

Similarly, the concept of the “KAP gap” states that even if an individual is introduced to new ideas or practices and has a positive attitude toward these new ideas or practices, the individual will not necessarily adopt the behavior [41]. This disconnect results in inconsistency and creates a sense of cognitive dissonance. This problem is especially apparent with respect to new preventive ideas or practices because they often lack positive benefits and advantages. Thus, the effect is more ambiguous and individuals are more prone to experiencing a KAP gap. This tendency may explain why medical majors and those who frequently engage in health discussions do not necessarily implement their knowledge into actions. Future studies are needed to examine the motivation for one’s decision to adopt, reject, or discontinue a healthy behavior (ie, what leads to the KAP gap).

Partially consistent with IMeHU [9], we found that critical health literacy played a key role in health status and health behaviors. Functional health literacy positively predicted eating and exercise behaviors, and critical health literacy positively predicted the 3 dimensions of health behaviors. However, interactive health literacy did not contribute to the 3 dimensions of health behavior. Accordingly, hypothesis 2 was partially supported. Individual health information literacy strengthens one’s intent to search for and apply health information and influences one’s health decision making and engagement in healthy behaviors [18]. Thus, the current study concludes that both functional and critical health literacy and dialog contribute to the adoption of positive health behaviors, especially eating and exercising behaviors.

However, functional eHealth literacy was not found to affect sleep behaviors, and interactive eHealth literacy was not found to affect any of the health behavior dimensions. Therefore, it is inferred that the factors that influence health behaviors are complex. The promotion of individual eHealth literacy affects the retention of eHealth information and subsequently influences future actions [9]. According to involvement theory [42], critical eHealth literacy may motivate individuals to seek and evaluate the quality of health information. In other words, individuals may attempt to gain as much information as possible and then evaluate and use the information to reach an optimal decision. The processing involved in functional and interactive eHealth literacy does not engage as deeply with an issue as that of critical eHealth literacy. Thus, functional and interactive eHealth literacy are less influential than critical eHealth literacy.

The findings also revealed that functional and critical eHealth literacy mediated the association between individual factors and health behaviors. Thus, hypothesis 4 was partially supported. We found that functional eHealth literacy only mediated the association between health concern and 2 health behavior dimensions (eating and exercise behaviors). Critical eHealth literacy mediated not only the association between health concern and eating behaviors, but also the association between health status and all 3 health behavior dimensions. Therefore, college students’ critical eHealth literacy influenced the association between perceived health status and the 3 health behavior dimensions. These findings are consistent with the belief that critical literacy is crucial to assess the quality of health information because laypersons risk harming themselves in their self-diagnosis and treatment when they lack the required background knowledge to correctly interpret information [43].

This study used the IMeHU to explore the associations among individual factors, eHealth literacy, and health behaviors. We hope that the findings will stimulate further debate about how a health education framework can be translated into practical approaches and will contribute to further refinement of the eHealth literacy concept. The study showed that eHealth literacy played an intermediary role in the association between individual factors and health behaviors. Thus, we suggest that schools strengthen college students’ eHealth literacy and promote positive health behaviors based on the current level of eHealth literacy among students [44]. Schools can further use the 6 core skills of eHealth literacy [45] (eg, traditional literacy, health literacy, information literacy, scientific literacy, media literacy, and computer literacy) to develop healthy eating, exercising, and sleeping behavior guidelines and to incorporate these guidelines into health education programs. Moreover, the current findings demonstrated that critical health literacy is a key competence in promoting individual health behaviors. Thus, it is suggested that the development of critical eHealth literacy and the promotion of positive health behaviors among college students require further investigation.

As this study indicated, the IMeHU [9] provides a parsimonious framework that is currently lacking in the extant literature. The examination of the potential effect of eHealth literacy on health behavior presents a unique challenge because it involves a complex interplay of basic literacy skills, the ability to successfully navigate the dominant language framework (English) and culture that is utilized for Web-mediated communication, and sufficient levels of technology adoption and proficiency [46]. Although some research hypotheses failed to gain support in this study, given that the factors that influence health behaviors are complex and interdependent, a future study should be conducted to explore how these factors influence one another. Further research is also needed to refine and verify the parsimonious framework of the IMeHU [9].