COVID‐digital health literacy and subjective well‐being of students in Ghana: Mediation‐moderation analyses

Abstract Background Previous research has established a strong association between COVID‐19 digital health literacy (DHL) and subjective well‐being among several populations, including students. With the growing misinformation and heightened fear of COVID‐19 among persons with an underlying medical condition, several scholars have questioned the direct relationship between DHL and well‐being. This study assessed the moderating roles of information accuracy concerns and the existence of an underlying medical condition among students. Methods Using a cross‐sectional design, a multi‐stage sampling approach was used to select 1392 students from senior high schools in Northern Ghana who completed a questionnaire containing information on DHL, information accuracy, subjective well‐being, and underlying health condition, with reported internal consistency coefficients above 0.70. The data which was processed with SPSS version 25, was analyzed using correlation (Pearson and biserial), and Hayes' PROCESS for the moderation and mediation analyses. Results A significant positive relationship was found between (a) DHL and subjective well‐being, (b) DHL and information accuracy concerns, and (c) information accuracy concerns and subjective well‐being. However, the prevalence of underlying health condition was negatively associated with information accuracy, DHL, and subjective well‐being. Information accuracy concerns and the existence of an underlying medical condition significantly regulated the relationship between DHL and subjective well‐being. Conclusions Demonstrating satisfactory levels of DHL does not necessarily result in improved subjective well‐being. However, emphasis should be placed on whether individuals attach much importance to the accuracy of information retrieved as well as having or not an underlying health condition.


| INTRODUCTION
The outbreak of COVID-19 pandemic has affected the lives and mental health of students globally. [1][2][3][4][5][6][7] These psychological consequences are related to the spread of information and misinformation (infodemic) about this epidemic owing to the prevalence of digital social media as the main source of information search among the general public, including adolescents and young adults. 8,9 Recent studies have found that young adults showed a heightened use of digital media during the COVID-19 pandemic. 10,11 Subsequently, digital health literacy (DHL) has become a public health concern. 12,13 DHL refers to the skills to search, select, understand, appraise online health information and health care and apply the knowledge gained for preventing, addressing, or solving a health problem to improve psychological well-being. [14][15][16][17][18][19][20] A substantial literature has indicated that students have reported moderate to high levels of COVID-DHL (i.e., sufficient DHL) in Germany, 20 Portugal, 22 Spain and France, 23 Slovenia, 24 and Denmark. 12 Conversely, a low level of COVID-DHL has been also reported among students in America 9 , and Europe. 25 Extant researchers have discovered that higher DHL is associated with subjective well-being among young adults. 19,26,27 For example, in China, a positive association between COVID-DHL and the psychological well-being of students has been reported. 15,28 Similarly, Nguyen et al. 16 in Vietnam established that COVID-DHL positively influences the subjective well-being of students. These results suggest that strengthening students' DHL could empower them to actively partake in managing their health by accessing and using accurate health information.
The link between students' COVID-DHL and subjective wellbeing could be explained by the information accuracy screened by the students. However, this relationship has not yet been tested.
Information accuracy concern is the extent to which students are concerned about the correctness of the information they search online about coronavirus. Identifying which information is accurate and helpful requires DHL. 26,29 Accessibility to correct information increases COVID-19 health literacy. 8 Nguyen et al. 16 in Vietnam found a significant positive association between the importance of online information content related to COVID-19 and the subjective well-being of students, while Zakar et al. 13 in Pakistan observed a negative association between students' subjective well-being and information accuracy. Other researchers have also found the importance of information to be positively related. 13,16 Based on these results, we assume that information accuracy could mediate the relationship between DHL and the subjective well-being of students.
Students with "underlying health condition" are at greater risk of COVID-19 which could affect their subjective well-being. Accordingly, these students need a level of COVID-DHL to be able to evaluate COVID-19 information for use in their daily lives. Also, "underlying health condition" could affect the strength and direction of the relationship between students' COVID-DHL and subjective well-being. This relationship has also not yet been tested. However, some studies have established a link between health literacy and health condition. [30][31][32][33][34] Holt et al. 34 for example, found that students with an underlying health condition demonstrated lower levels of DHL in Denmark. Liu et al. 30 in China, on the other hand, found that persons with adequate health literacy are more likely to have reduced chronic disease. Also, in a review, van der Heide et al. 32 found that many studies reported lower health literacy to be associated with a high occurrence of chronic diseases. Gleaning from the findings of these previous studies, DHL and underlying health condition are related. Likewise, DHL and subjective well-being. Therefore, there is a likelihood that the underlying health condition would moderate the relationship between COVID-DHL and subjective well-being.
In Ghana, digital divide is a major concern among the general public, including young adults and students. 1,35 The adolescents (including students) in Ghana could have low DHL, which could partly be explained by low socioeconomic and poor health status. Most of the students in Ghana are frequent users of social networks.
However, due to the influx of information from varied sources, extracting useful data remain a challenge. Adolescents are not fully equipped to critically analyse information and take appropriate personal and community health decisions.
Overall, there is a paucity of evidence with reference to COVID-DHL and subjective well-being among students in Ghana despite a large volume of literature on these phenomena across western countries. 15,16,26 The study assessed the (a) linear relationship between the pairs of the following variables: DHL, subjective wellbeing, information accuracy and prevalence of underlying health conditions, (b) mediating role of information accuracy in the relationship between DHL and subjective well-being, and (c) moderating effect of the prevalence of underlying health condition in the relationship between DHL and subjective well-being. The outcome of this inquiry would provide useful information on the vulnerability of students during crises and inform policymakers to provide health educative interventions in schools aimed at strengthening DHL among students to increase their subjective well-being.

| Study participants
This study was cross-sectional survey, with data gathered from 1392 senior high school (SHS) students from Northern and the two upper regions of Ghana using a multi-stage sampling approach. Through a random sampling procedure, the Upper West and Savannah regions were first selected. Five schools were then targeted from each region using a cluster sampling technique. This step was followed by purposefully sampling students from each school, considering whether the participants had stayed and/or schooled within the Northern zone of Ghana for more than 10 years. Using G-power (Version 3.1.9.2) with the actual power of 0.950, seven predictors (including moderation and mediation models), and F-tests (linear multiple regression: fixed model, F 2 deviation from zero), a sample size of 1099 was obtained. Furthermore, 30% of the estimated sample size yielded a total sample of 1429. However, 37 responses were not retrieved, leading to a response rate of 97.4%. In the study, 702 (50.4%) and 654 (47.7%) were males and females respectively.  36 In this study, the recommendations from a recent revalidation of DHLI 37

| Subjective well-being
The subjective well-being of the participants was measured using the World Health Organization (WHO) well-being index. 38 The WHO-5 index is a unidimensional scale with responses from 0 to 5 (with 0 representing "at no time," 1 signifying "some of the time," 2 indicating "less than half of the time," 3 representing "more than half of the time," 4 being "most of the time," and 5 being "all of the time"). The properties of the WHO-5 index are sufficient. 39 The reliability coefficient estimate (Omega ω) reported in this study was 0.815. The sum score was computed for all items and then converted to 100 for use in the follow-up analyses.

| Information accuracy concerns
The information accuracy concerns was conceptualized as the extent to which the participants were concerned about the correctness of the information they search online about coronavirus. The instrument was developed by Gebel et al. 40 with the following preamble: "This section is about how important various things are to you when you search the internet about coronavirus and related topics. How important is it to you that……". The instrument contained 6-items, which are: "the information is verified", different opinions are represented", "you quickly learn the most important things", "the subject is dealt with comprehensively", "the information is up to date." The items had responses measured on a 4-point scale (0-not at all important, 1-rather not important, 2-rather not important, and 3very important). A reliability estimate (Omega ω) of 0.771 was reported in this study which was deemed adequate. 41 The sum score was computed for all items, and this was used in the follow-up analyses.

| Prevalence of underlying health condition
The 'underlying health condition' variable sought to measure the prevalence of medically diagnosed conditions which makes people vulnerable to coronavirus disease. 42 The participants in this study were asked to indicate if they had any chronic underlying health conditions. These conditions include lung disease, obesity, cancer, liver disease, diabetes mellitus, neurological disability, mood health disorder, and tuberculosis, among others. If a respondent had any one of the following conditions, a "yes" response was provided whereas a 'no' response indicated the absence of any chronic health condition. 43 This variable was treated as categorical.

| Procedure
Ethical approval was obtained from the University of Education, Winneba Ethical Review Board with a reference number DAA/ P.1/Vol.1/39. In addition, the headmasters of all the senior high schools in the Northern and Upper regions that participated in this particular study gave their approval. Without considering the tribe a student belonged to, the researchers recruited SHS students who had stayed in the Northern and Upper regions for more than 10 years or come from any of the Northern towns.
Students who can eloquently speak, read, understand, and write English were used for the study.
After the researchers spoke with 12 research assistants who voluntarily agreed to help in the data collection process, they were taken through the questionnaire from the beginning to the end where each instruction and item were thoroughly explained to them. To facilitate the data collection process, a good rapport was created between the researchers, the research assistants, and members of the school community (headmaster, teachers, and students) through a visit to the selected schools. The researchers took advantage of the visit to explain the rationale of the study to both students and teachers. The researchers then recruited students who were accessible, ready, and willing to be part of the research and were QUANSAH ET AL. told could opt out any time they willed without any problem.
Confidentiality and anonymity were assured, with all COVID-19 safety protocols observed.
Provision was made for nose masks, hand sanitizers, water, and liquid soap for hand washing, and tissue papers for cleaning wet hands to ensure that students were not exposed to COVID-19 infection. The questionnaire was given to students to answer during the free periods on the schools' timetable after complying with all ethical issues. It took students 15-20 min to respond to items on the questionnaire. All answered items were taken and kept safely by the researchers.

| Statistical analyses
The data for DHL, subjective well-being and information accuracy concerns were normally distributed. The data were analyzed with different statistical procedures in line with the objectives of the study. First, a correlational analysis (using SPSS version 25) was conducted to establish how the variables (i.e., DHL, subjective well-being, information accuracy, and prevalence of underlying health conditions) in the study were related. While Pearsonproduct moment correlation was used for relationships among DHL, subjective well-being, and information accuracy, a biserial correlation was used for each of the aforementioned and prevalence of underlying health conditions. Correlation analysis was performed because the focus was to determine bidirectional associations among the variables. Secondly, a simple mediation analysis was performed using the Hayes PROCESS add-on macro, 44 to assess the mediating effect of information accuracy in the relationship between DHL and subjective well-being.
Specifically, the Model 4 option was used with multiple predictors using 5000 bootstrap samples. Hayes PROCESS was used for the mediation analysis because it is one of the current ways of testing indirect effect which has shown to be robust in detecting significant indirect effect relative to previous methods such as the Sobel test and Baron and Kenney's approach. Additionally, its use of bootstrap samples helps in estimating unbiased indirect effect with minimal errors. Lastly, simple moderation analysis was also performed with Hayes PROCESS add-on macro 44 Table 1 presents the distribution of the demographic characteristics of the respondents, which include sex, age range, class, and whether they had any underlying condition.

| Demographic profile of respondents
There were more male than female respondents, with male students constituting 50.4% of the sample and females representing 47%. About 2.6% of the respondents identified themselves as diverse (see Table 1 Although most of the students indicated that they had no underlying chronic condition (76.1%), quite a number of them also reported having an underlying chronic condition.

| Relationship among the DHL, subjective well-being, information accuracy, and prevalence of underlying health conditions
The study explored the relationship existing among the study variables as well as their descriptive statistics as presented in  Table 2). Information accuracy concerns was also positively related to DHL (r = 0.538, p < 0.001) and subjective well-being (r = 0.432, p < 0.001). It was further found that respondents with underlying chronic health conditions reported lower levels of DHL (r = −0.470, p < 0.001) and subjective wellbeing (r = −0.317, p < 0.001).

| Mediating role of information accuracy concerns in relationship between digital health literacy and subjective well-being
The study also assessed the mediating role of information accuracy concerns in the relationship between DHL and subjective well-being.
A mediation analysis was performed using subjective well-being as a criterion variable, information accuracy concerns as a mediator, and the dimensions of DHL as the predictors (i.e., searching for information, adding self-generated information, assessing information reliability, and determining information relevance). The analysis details are shown in Table 3.
The results, as presented in Table 3, showed that information accuracy concerns significantly mediated the relationship between DHL and subjective well-being, β = 1.337, BootSE = 0.595, BootCI

| Moderating role of underlying health condition in the link between digital health literacy and subjective well-being
The study additionally sought to establish the moderating role of underlying health conditions in the relationship between DHL and subjective well-being. Performing a moderation test, DHL served as the predictor, subjective well-being acted as the criterion variable and the prevalence of underlying health conditions served as the moderator variable. The details of the analysis are presented in Table 4.
As presented in Table 4 Further probing analysis was conducted by generating and inspecting the moderation curves to understand the nature of the moderation effect (see Figure 1). It was discovered that, for those without underlying health condition, the relationship between DHL and well-being was steeper (strong) than for those with medical health conditions (weak). In other words, with the same level of DHL, students without health conditions were more likely to experience sufficient levels of well-being.

| DISCUSSION
One of the objectives of this study was to explore the relationships between DHL, subjective well-being, information accuracy, and prevalence of an underlying health condition among students in  information obtained through the use of digital media. This assertion is pertinent considering the heightened fear associated with the pandemic resulting in its infodemic, 45,46 and its impact on the well-being of students. [47][48][49] Previous studies attributed this infodemic to the influx of digital social media. 8,9 It is noteworthy that, in the case of self-generated content as a Overall, information accuracy mediated the relationship between DHL and subjective well-being. It must be acknowledged that though previous studies have not examined the intermediary role of information accuracy, some have found a link between DHL and information accuracy. 8,26 Similarly, the link between information accuracy and subjective well-being has been established. 13,16 It can be gleaned from these findings a possible mediating role of information accuracy, which supports the findings of this current study. Therefore, it is reasonable to state that DHL, solely, is not a panacea to poor or improved subjective well-being among students.
However, showing concern for the accuracy of health information obtained as well as the exhibition of high ability in the search and evaluation of health-related information appears to impact subjective well-being better.
The study further examined the moderating role of underlying health condition in the relationship between DHL and subjective well-being, which a significant moderation was revealed. Practically, the moderating effect was large.

| Limitations
The study was limited to students in senior high schools in Northern

| CONCLUSION
The study stresses that demonstrating satisfactory levels of health digital literacy does not necessarily result in improved subjective well-being.
However, much emphasis needs to be placed on whether individuals attach much importance to the accuracy of information retrieved as well as having or not having underlying health conditions. Intervention programs should be directed to students who (a) do not see the need for being concerned about the accuracy of the information they search for and (b) have an underlying medical health condition.

ACKNOWLEDGEMENTS
The study received no external funding. However, the authors sincerely thank Bielefeld University, Germany for providing financial support through the Institutional Open Access Publication Fund for the article processing charge (APC). Open Access funding enabled and organized by Projekt DEAL.

CONFLICTS OF INTEREST
The authors declare no conflicts of interest.

DATA AVAILABILITY STATEMENT
Anonymized data is available upon reasonable request through the corresponding author and takes complete responsibility for the integrity of the data and the accuracy of the data analysis.

INSTITUTIONAL REVIEW BOARD STATEMENT
The study was approved by the Institutional Review Board of the the University of Education, Winneba, Ghana with document number DAA/P.1/Vol.1/39.

INFORMED CONSENT STATEMENT
Written informed consent was taken from all study participants before data collection.

TRANSPARENCY STATEMENT
The lead author John Elvis Hagan affirms that this manuscript is an honest, accurate, and transparent account of the study being reported; that no important aspects of the study have been omitted; and that any discrepancies from the study as planned (and, if relevant, registered) have been explained.