Convergent and discriminant validation by the multitrait-multimethod matrix.

Abstract

<p>This data set was collected to examine the eHealth Literacy Scale as a means of assessing eHealth literacy among German athletes. In this context, we pursued two objectives with this study: first, to test the factorial structure of the GR-eHEALS and assess its construct validity by examining both convergent and discriminant validity; and second, to examine the associations between eHealth literacy and health-related outcomes (i.e., substance use and injuries). This digital survey was conducted as a cross-sectional study, adhering to the approval guidelines of the Ethics Committee of the Faculty of Medicine at the University of Duisburg-Essen (19-8947-BO). Prior to the survey, each participant provided electronic informed consent. Participation was both anonymous and voluntary, without any form of reimbursement. We utilized the Unipark software (Tivian XI GmbH) that was distributed through social media, sports clubs (involving athletes competing in regional and nationwide tournaments), and sports associations (both regional and nationwide) from December 2021 to December 2022.</p> <b>Methods</b> <p>This study collected sociodemographic data from participants through self-report measures, including information on their sex, age, marital status, education level, occupation, and financial situation. Furthermore, sports-related data (i.e., type of sports and whether they do individual or team sports) was assessed. eHealth literacy of participants was assessed using the GR-eHEALS (1), which is based on the eHEALS by Norman and Skinner (2). The GR-eHEALS consists of eight items that are rated on a 5-point Likert scale (1 = do not agree at all; 5 = fully agree).</p> <p>To test the convergent validity of the GR-eHEALS, established scales measuring confidence in using digital media (3, 4) were administered. Furthermore, the length of daily internet use for personal and professional purposes was evaluated using a single self-developed item rated on a 5-point Likert scale (1 = not at all; 5 = more than 5 hours). In addition, three items each were inquired after internet anxiety and digital overload (3, 5), and they were rated on a 5-point Likert scale (1 = totally disagree; 5 = totally agree). These measures were expected to correlate significantly with the GR-eHEALS, as per Campbell and Fiske (6) guidelines. To evaluate the discriminant validity of the GR-eHEALS, we used the 8-item Impulsive Behavior–8 Scale (7) to measure impulsivity as a personal trait that was expected to be independent of eHealth literacy. These items were also rated on a 5-point Likert scale.</p> <p>Additionally, participants provided medical data through self-report measures. It was assessed how often the following substances were consumed on a 5-point Likert scale (1 = never, 5 = daily): Cannabis, nicotine, sedatives prescribed by physicians (e.g. benzodiazepines), painkillers prescribed by physicians (e.g. tramadol), sedatives not prescribed by physicians / over-the-counter sedatives, painkillers not prescribed by physicians / over-the-counter painkillers (e.g. ibuprofen, diclofenac). Moreover, number and severity of injuries was assessed. For this purpose, athletes indicated on a 5-point Likert scale (1 = not at all, 5 = more than 20 times a year) how often they had suffered minor, moderate and severe injuries within the last year and how often surgery had been necessary.</p> <i>Statistical analyses</i> <p>Statistical analyses were conducted with R version 4.2.2.2 and R Studio 2023.06.1+524. A confirmatory factor analysis (CFA) was performed in order to affirm the factor structure of the GR-eHEALS scale in the present sample. Results were interpreted according to Hu and Bentler (8) assuming the comparative fit index (CFI) and Tucker Lewis index (TLI) of at least 0.95 and root mean square error of approximation (RMSEA) and standardized root mean square residual (SRMR) of below 0.06 and 0.08, respectively (8). As the GR-eHEALS consists of items on ordinal scale, a robust likelihood estimator (WLSMV; 9) was chosen to avoid biases in the model. Internal consistencies (reliability) of the convergent and discriminant validity scales, the GR-eHEALS and its two subscales were examined. Subsequently, two-tailed Pearson correlations were conducted between the validity scales, the outcome measurements, and sociodemographic variables with the GR-eHEALS. Sex differences on GR-eHEALS were assessed by a two-tailed independent t-Test. Results were considered as significant with p = .05. Incomplete data was deleted list wise.</p>