Relationship Between Electronic Health Literacy and Self-Management in People With Type 2 Diabetes Using a Structural Equation Modeling Approach

ABSTRACT Background Electronic health (eHealth) literacy is a relatively new concept used to determine health outcomes. However, it is not well known how eHealth literacy relates to health outcomes such as diabetes self-management. Purpose This study was designed to examine the relationships among eHealth literacy, self-efficacy, social support, and self-management in people with Type 2 diabetes. Methods A cross-sectional design was used to examine secondary data from a field survey of people with Type 2 diabetes recruited from outpatient clinics from August to December 2021 (N = 453). A structural equation model was used that first analyzed the measurement model using confirmatory factor analysis and then tested the hypothesized structural model to estimate the expected relationships among the study variables. The significance of the statistical estimates for the model was assessed based on the 95% bias-corrected bootstrap confidence interval from 5,000 bootstrap resamples. Results Significant, indirect relationships were found between eHealth literacy and self-management via self-efficacy (β = 0.26, B = 0.17, 95% CI [0.10, 0.24]) and via social support and, in turn, self-efficacy (β = 0.08, B = 0.05, 95% CI [0.04, 0.08]). eHealth literacy, social support, and self-efficacy together explained 58.1% of the variance in self-management. Conclusion/Implications for Practice This study provides new evidence regarding how eHealth literacy relates to self-management in people with Type 2 diabetes via two indirect pathways, including self-efficacy alone and social support and self-efficacy in series. An eHealth literacy program for self-management should be developed in clinical practice that includes strategies for inducing synergistic effects from self-efficacy and social support on self-management in people with Type 2 diabetes.


Introduction
As of 2022, an estimated 62% of the global population were active internet users (Statista Research Department, 2022), with most (76.9%) using the internet as a source of health information and services (Bujnowska-Fedak et al., 2019).Health information can be accessed more conveniently, cheaply, and quickly via the internet using digital devices (e.g., computers, tablets, and smartphones) than via traditional methods (e.g., visiting health professionals, attending health education programs, reading books).However, there are large amounts of information, misinformation, and disinformation circulating on the internet, particularly on social media platforms (Liu & Xiao, 2021).This means that individuals need to be able to identify accurate and trustworthy health information on the internet, which is an ability called electronic health (eHealth) literacy.
The term "eHealth literacy" first appeared in the literature in 2006, when it was defined as "the ability to seek, find, understand, and appraise health information from electronic sources and apply the knowledge gained to addressing or solving a health problem" (Norman & Skinner, 2006).Studies on eHealth literacy began in earnest after 2014, exploring the associations between eHealth literacy and health outcomes such as lifestyle behaviors and self-management (Wang et al., 2022).However, these studies have been scarce and have obtained unclear findings based on inadequate evidence (Karnoe & Kayser, 2015;Neter & Brainin, 2019).For example, some researchers have reported weak correlations between eHealth literacy and health behaviors or self-care in college students (Hwang & Kang, 2019) and patients with heart failure (Chuang et al., 2019).However, others have found no correlation with self-care in young adults (Nugroho et al., 2021) or in patients with hypertension (Cho & Ha, 2019).Han et al. (2018) also found the Methods: A cross-sectional design was used to examine secondary data from a field survey of people with Type 2 diabetes recruited from outpatient clinics from August to December 2021 (N = 453).A structural equation model was used that first analyzed the measurement model using confirmatory factor analysis and then tested the hypothesized structural model to estimate the expected relationships among the study variables.The significance of the statistical estimates for the model was assessed based on the 95% bias-corrected bootstrap confidence interval from 5,000 bootstrap resamples.

Conclusion/Implications for Practice:
This study provides new evidence regarding how eHealth literacy relates to self-management in people with Type 2 diabetes via two indirect pathways, including self-efficacy alone and social support and self-efficacy in series.An eHealth literacy program for self-management should be developed in clinical practice that includes strategies for inducing synergistic effects from self-efficacy and social support on self-management in people with Type 2 diabetes.
associations between eHealth literacy and health outcomes in people with HIV to be inconsistent and suggested the need for further studies with more rigorous methodologies (e.g., adequate sample sizes and the use of validated eHealth literacy measures).
Diabetes is a worldwide health issue.According to the International Diabetes Federation ( 2021), approximately 537 million adults currently live with diabetes (10% of the population), and this number is expected to increase to 643 million by 2030.Type 2 diabetes accounts for more than 90% of diabetes cases in adults worldwide.Self-management is vital for people with Type 2 diabetes to optimize metabolic control and prevent acute and chronic complications.Under self-management, patients are expected to achieve their goals by making day-to-day decisions about physical exercise, food, medication adherence, and blood glucose self-monitoring.Diabetes self-management requires appropriate information and skills, and the internet appears to be the source most frequently used by these patients for diabetes information (Kuske et al., 2017).People with high eHealth literacy levels appear to be able to improve their health outcomes by using information obtained from the internet (Neter & Brainin, 2019).eHealth literacy is therefore expected to be an essential self-management skill for people with diabetes living in the current internet era (Sjöström et al., 2021).However, the direct relationship between eHealth literacy and self-management was empirically found to be very weak in people with Type 2 diabetes (K. A. Kim et al., 2018).In light of this, the association was suggested to be indirect and affected by potential mediators (Xie et al., 2022).If this is the case, it is important to determine the mediators that explain the mechanism underlying the link between eHealth literacy and self-management.
According to social cognitive theory, self-efficacy refers to the beliefs of individuals in their capabilities to attain certain goals (Bandura, 1977).Although little evidence exists, caregivers and college students with higher eHealth literacy levels have been considered more likely to have higher self-efficacy levels (Efthymiou et al., 2022;S. Kim & Jeon, 2020).People with higher self-efficacy levels perceive their tasks or behaviors as challenges to be achieved and thus tend to engage in them successfully (Bandura, 1977).Self-efficacy was therefore deemed a crucial predictor of diabetes self-management by the Association of Diabetes Care and Education Specialists and Kolb (2021) as well as by other researchers (E.-H. Lee et al., 2016).Thus, in this study, eHealth literacy is postulated to have an indirect relationship with self-management via a mediating effect from self-efficacy.
Social support is the self-perceived availability of resources provided by a network that encompasses an individual and consists of family members, friends, and professionals.These resources may be instrumental (tangible goods or services), informational (education or advice), appraisal (affirmation), or emotional (affection; Langford et al., 1997).In previous studies, eHealth literacy was found to relate positively with social support in patients with congestive heart failure (Chuang et al., 2019) and in primary health providers (Xu et al., 2022).
Social support has been generally considered to enhance diabetes self-management (Al-Dwaikat & Hall, 2017).Therefore, it is proposed in this study that eHealth literacy may be indirectly related to self-management via a mediating effect from social support.
However, some researchers have found empirically that the direct relationship between social support and self-management became insignificant when the mediating effect of self-efficacy was included (Chan et al., 2020;Shao et al., 2017).They also noticed an indirect (rather than direct) relationship between social support and self-management via self-efficacy.By also considering eHealth literacy, it was hypothesized in this study that eHealth literacy is related indirectly to self-management via social support and, in turn, self-efficacy.
The aim of this study was to examine the relationships among eHealth literacy, self-efficacy, social support, and self-management in people with Type 2 diabetes.This study is the first to the authors' knowledge in which these variables were empirically examined together.On the basis of the aforementioned literature, the following hypothesis was proposed to describe the link between eHealth literacy and self-management: eHealth literacy relates directly to self-management, self-efficacy, and social support; self-efficacy relates directly to self-management; social support relates directly to self-efficacy and self-management; eHealth literacy relates indirectly to self-management via self-efficacy; eHealth literacy relates indirectly to self-management via social support; and eHealth literacy relates indirectly to self-management via social support and, in turn, self-efficacy.

Study Design and Sample Size
This study used a cross-sectional design based on structural equation modeling (SEM) using secondary data from a field survey for the development of an instrument for people with Type 2 diabetes who were recruited from outpatient clinics from August to December 2021 (E.-H. Lee et al., 2022).The data collection process and participant information are presented in detail elsewhere (E.-H. Lee et al., 2022).Women comprised 35.3% of the 453 participants.The mean age was 56.8 years (SD = 10.8 years), and 46.4% had graduated from college.About two thirds of the participants (78.1%) were taking an oral hypoglycemic agent, and one third had controlled HbA1c levels (≤ 6.5%).The sample size of 453 in this study satisfied the requirement of being at least 10 times larger than the number of estimated parameters (Kline, 2010).

Instruments
Electronic health literacy eHealth literacy was assessed using the Condition-Specific eHealth Literacy Scale for Diabetes (CeHLS-D) that was developed recently to reflect the social nature of Web 2.0 (E.-H. Lee et al., 2022).This scale comprises 10 items scored on a The Journal of Nursing Research Eun-Hyun LEE et al.
5-point Likert scale, ranging from 0 (not at all) to 4 (very much), in two subscales: (a) cognitive actions for internet diabetes information (e.g., "thinking of search words" and "appraising information credibility") and (b) digital communication ability (e.g., "text messaging").Higher CeHLS-D scores indicate higher eHealth literacy levels.The CeHLS-D exhibited good psychometric properties in terms of internal consistency (Cronbach's alpha for subscales = .92-.89, omega for subscales = .94-.87) and content, structural, convergent, and known-groups validities among people with Type 2 diabetes.Measurement invariance was also satisfied across gender, age, and glycemic control groups.The CeHLS-D had a Cronbach's alpha of .94 in this study.

Self-efficacy
The Korean version of the Diabetes Management Self-Efficacy Scale (DMSES; E.-H. Lee et al., 2015) was used.The DMSES consists of 16 items on an 11-point scale, with higher scores representing greater self-efficacy for diabetes self-management.This scale comprises four subscales: nutrition, physical exercise, medical treatment, and blood glucose.Total/subscale scores are calculated as the sum of the scores for each total/subscale item.The 16-item scale exhibits good psychometric properties in terms of content, structural, and concurrent validities; internal consistency (Cronbach's alpha = .92);and test-retest reliability (intraclass correlation coefficient = .85).The DMSES earned a Cronbach's alpha of .94 in this study.

Social support
Social support was measured in this study using the modified Medical Outcomes Study Social Support Survey (mMOS-SS; Moser et al., 2012).The mMOS-SS comprises eight items scored on a 5-point Likert scale, with higher scores indicating greater social support.This instrument has two subscales (instrumental and emotional) and has exhibited good psychometric properties in terms of structural, construct, and divergent validities and in terms of internal consistency (Cronbach's alpha = .88-.93 among three populations).The mMOS-SS earned a Cronbach's alpha of .96 in this study.

Self-management
Self-management was measured in this study using the Diabetes Self-Management Scale, which comprises 17 items covering two components (lifestyle behaviors and regimen behaviors) and is scored on a 5-point Likert scale, with higher scores indicating better self-management (E.-H. Lee et al., 2020).This scale exhibited satisfactory content, structural and convergent validities, internal consistency (Cronbach's alpha = .76-.84), and test-retest reliability (intraclass correlation coefficient = .84-.92) among 473 people with Type 2 diabetes.The Diabetes Self-Management Scale earned a Cronbach's alpha of .88 in this study.

Data Analysis
Data were analyzed using SPSS Version 25 and AMOS software run on Windows (IBM, Inc., Armonk, NY, USA).The means and SDs of the study variables were computed using descriptive statistics.The skewness and kurtosis of the variables were calculated as absolute values, with values less than 2 and 7, respectively, considered indicative of univariate normality (Hair et al., 2014).Multicollinearity was examined using the variance inflation factory.The bivariate correlations among the study variables were analyzed using Pearson's correlation coefficients.The internal consistencies of the instruments were evaluated using Cronbach's alpha.
SEM with a two-step approach was used to test the hypothesized model (Hair et al., 2014).In the first step, the measurement model was analyzed using confirmatory factor analysis with maximum likelihood to identify whether indicators (observed variables) loaded on latent variables as proposed.The overall goodness of fit for the measurement model was evaluated based on the following fit indices and other criteria: comparative fit index (CFI) > .95,standardized root-mean-square residual (SRMR) < .09,and root-meansquare error of approximation (RMSEA) < .10(Byrne, 2016).The traditional w 2 value and degree of freedom were reported but not used to determine the model fit, as they are sensitive to sample size (Hair et al., 2014).A standardized factor loading value higher than .50 was considered indicative that the indicator (observed variable) had converged on the latent variable (Hair et al., 2014).A heterotrait-monotrait ratio of correlations of < .85 was considered indicative that a pair of latent variables was discriminant (Henseler et al., 2015).
In the second step, the hypothesized structural model was tested to estimate the expected relationships among the latent variables.The overall goodness of fit of the structural model was evaluated using the abovementioned indices (SRMR, CFI, and RMSEA).The significances of the individual parameter estimates for the structural model were determined using the critical ratio and a criterion of p < .005.The model was modified when a nonsignificant parameter was found.Bayesian information criterion (BIC) values were used to select the optimal model (Byrne, 2016).The selected optimal model was also double-checked using the function of the model specification search in AMOS.The significance of statistical estimates for indirect effects was assessed based on the 95% bias-corrected bootstrap confidence interval (CI) from 5,000 bootstrap resamples.If the CI did not include zero, the indirect effect was considered significant.Because the default bootstrap procedure in AMOS obtains the total indirect effect, the specific indirect effects of eHealth literacy on diabetes self-management were estimated and statistically determined using phantom variables.Finally, the squared multiple correlations of the self-management variables in the structural model were calculated.

Ethical Considerations
This study was approved by the institutional review boards of the hospitals (Approval Numbers AJIRB-MED-SUR-21-179 and INHAUH 2021-07-022).All of the participants signed a written informed consent form.

Descriptive Statistics and Correlations Among Study Variables
The mean scores for eHealth literacy, self-efficacy, social support, and self-management were 2.39 (SD = 1.04), 108.98 (SD = 29.58),2.44 (SD = 1.10), and 2.16 (SD = 0.70), respectively.The absolute values for the skewness (0.01-0.67) and kurtosis (0.18-1.34) of the study variables met their respective normality criteria.As the variance inflation factor values were less than 10, no evidence of high multicollinearity was found.In addition, significant correlations were found among the study variables (Table 1).

Measurement Model
The measurement model with four latent variables provided a marginal fit to the data: w 2 = 263.83(df = 29, p < .001),CFI = .91,SRMR = .05,and RMSEA = .13(90% CI [0.12, 0.15]).The model was modified based on the highest modification index value: The measurement errors of two observed variables (medical treatment and blood glucose) were connected by two-headed curved arrows (Figure 1).The modified measurement model provided a good fit to the data: w 2 = 141.70 (df = 28, p < .001),CFI = .96,SRMR = .05,and RMSEA = .095(90% CI [0.08, 0.11]).Compared with the initial measurement model, the CFI value difference (ΔCFI = .05) of this modified model satisfied the criterion of ΔCFI > .01,indicating this model provided a meaningful improvement (Byrne, 2016).The standardized factor loadings for all indicators were higher than .50,which satisfied the criteria.The heterotrait-monotrait ratio of correlation values between latent variables were all less than .85(.32-.75).In summary, the latent variables in the model were well measured by their indicators (observed variables).

Structural Model
The overall model fit statistics indicate that the hypothesized structural model provided a good fit to the data: w 2 = 141.70 (df = 28, p < .001),CFI = .96,SRMR = .05,RMSEA = .095(90% CI [0.08, 0.11]), and BIC = 306.82.However, the coefficients of two pathways, namely, eHealth literacy to diabetes self-management and social support to diabetes self-management, were not significant.After eliminating the nonsignificant pathways, the modified model provided a good fit: w 2 = 144.88(df = 30, p < .001),CFI = .96,SRMR = .05,RMSEA = .09(90% CI [0.08, 0.11]), and BIC = 297.78(Figure 2).The BIC value of the modified structural model was smaller, indicating that this model provided a better fit (Fan et al., 2016).In addition, the model specification search function in AMOS indicated the modified structural model provided the best fit.Considering both model fit and parsimony, the modified structural model was selected as the final model for this study.
Regarding squared multiple correlations, eHealth literacy explained 11.5% of the variance in social support (R 2 = .12);eHealth literacy and social support together explained 28.6% of the variance in self-efficacy (R 2 = .29);and eHealth literacy, social support, and self-efficacy together explained 58.1% of the variance in self-management (R 2 = .58;Figure 2).

Discussion
A hypothesized model linking eHealth literacy to self-management, including the mediating effects of self-efficacy and social support in people with Type 2 diabetes, was investigated in this study.A bivariate analysis found a weak but significant correlation between eHealth literacy and self-management.This finding is consistent with that of a previous study (K. A. Kim et al., 2018).However, the direct relationship between   eHealth Literacy and Self-Management eHealth literacy and self-management was not supported when the mediating effects of self-efficacy and social support were included.This finding may reflect the effect of mediators on the relationship between eHealth literacy and health outcomes (Karnoe & Kayser, 2015).
The results of this study support an indirect relationship between eHealth literacy and self-management via self-efficacy.In other words, self-efficacy mediates the relationship between eHealth literacy and diabetes self-management.This finding is congruent with the findings of a previous study on patients with chronic disease (Wu et al., 2022).A systemic review conducted by Xie et al. (2022) emphasized the need for further exploration and elucidation of the various potential mediators of the relationship between eHealth literacy and health-related outcomes (including self-management).For example, Lin et al. (2020) identified psychological distress as a mediator in the relationship between eHealth literacy and medication adherence in patients with heart failure.In addition, eHealth literacy has been found to have a moderately strong association with perceived empowerment in adult internet users (Seçkin et al., 2016), and empowerment has been found to affect self-management (of diet and exercise) in people aged 60 years and older (Shin & Lee, 2018).In light of the above, further studies are necessary to examine psychological distress and empowerment in patients with diabetes as potential mediators in the relationship between eHealth literacy and self-management.
findings of this study support a direct relationship between eHealth literacy and social support, which is consistent with previous findings showing a relationship between higher eHealth literacy levels and greater social support (K. A. Kim et al., 2018;Xu et al., 2022).The findings in this study may be associated with individuals with higher eHealth literacy being able to better communicate and share information with each other in Web 2.0, which increases perceived social support.
An indirect relationship between eHealth literacy and self-management via social support was not supported in this study because the direct pathway from social support to selfmanagement, which exhibited a significant bivariate correlation in Pearson's analysis, was not significant in the hypothesized structural model.Thus, the condition for testing a simple mediation effect was not established because of the lack of a pathway between a mediator and a dependent variable.This may be because of social support being directly related to self-efficacy in the model.The finding of a nonsignificant direct pathway from social support to self-management was in agreement with other studies that included the mediating effect of self-efficacy (Chan et al., 2020;Shao et al., 2017).Similarly, Bandura (1977) asserted social support to be an external environmental source of self-efficacy that does not directly affect behavior.
The findings of this study supported an indirect relationship between eHealth literacy and self-management via social support and self-efficacy in series.This may be interpreted as individuals with better eHealth literacy being more likely to perceive that they have access to external resources, which in turn improves their confidence in their ability to successfully achieve their required tasks and engage in diabetes self-management.This pathway may be helpful for health professionals to obtain a better understanding of the mechanism that underlies the relationship between eHealth literacy and self-management in people with diabetes.
The two indirect relationships supported in this study suggest a future direction for clinical practice.When planning an intervention program for eHealth literacy to enhance diabetes self-management, strategies of improving self-efficacy and social support should be combined to induce a synergistic effect.This study also showed that the indirect effect of eHealth literacy on self-management via self-efficacy alone was stronger than the effect via social support and self-efficacy in series.This finding should be considered when planning intervention programs for people with Type 2 diabetes.

Strengths and Limitations
Most studies in the literature have measured eHealth literacy using the generic version of the Electronic Health Literacy Scale (Norman & Skinner, 2006).The Electronic Health Literacy Scale, a pioneering instrument for measuring eHealth literacy, was developed in the Web 1.0 era before the rise of social media and the mobile web.Thus, it has been criticized as being inadequate for measuring eHealth literacy given the social nature of Web 2.0 (J. Lee al., 2021).One strength of this study is its measurement of eHealth literacy using CeHLS-D, which was developed to reflect the nature of Web 2.0.Another strength is the examination of the mediation model using SEM, which has greater statistical power than the commonly used test proposed by Baron and Kenny (1986).A weakness of this study was its cross-sectional design, which disallows assessing temporal sequences in the mediation model.Thus, future studies should employ a longitudinal mediation model that, for example, includes analyses of autoregressive cross-lagged models or latent growth curve models (O'Laughlin et al., 2018).

Implications for Practice
Given the growing trend toward patient-centered care, eHealth services delivered via websites, apps, mobile apps, and similar virtual platforms have become important in delivering self-managed healthcare to patients with diabetes.Patients need to have sufficient eHealth literacy to effectively use eHealth services.Therefore, health clinicians must provide eHealth literacy programs to patients with diabetes to support their self-management efficacy.Strategies for self-efficacy and social support should be considered together to induce synergistic effects when providing such programs.However, the strategies for self-efficacy may be more focused than those for social support because the magnitude of the indirect effect of eHealth literacy on self-management via self-efficacy alone was larger than that of the serial effect via social support and self-efficacy in this study.

Conclusions
This study examined the direct and indirect relationships among eHealth literacy, self-efficacy, social support, and self-management in people with Type 2 diabetes.The findings yield new knowledge regarding how eHealth literacy relates to self-management in this vulnerable population.The relationship between eHealth literacy and self-management was shown to be linked by two indirect pathways, namely, via self-efficacy alone and via social support and self-efficacy in series.Future studies with longitudinal designs should be conducted to assess the temporal sequences of the indirect relationships that were found in this study.

Figure 2
Figure 2Modified Structural Model Linking eHealth Literacy to Self-Management

Figure 1
Figure 1 Modified Measurement Model

Table 2
Direct Pathways in the Modified Structural Model and Specific Indirect Effects Using Phantom Variables