226 7 Chapter 7 access to their digital health budget after completing all six questionnaires. Within the study demographic data, age, gender and educational background were collected. Educational level was categorised as low (no education to lowest high school degree), middle (vocational training to highest high school degrees) and high (university of applied sciences degree and research university degree). People were also asked to indicate whether they had a current medical, physical or psychological diagnosis. Data analysis Preparatory analyses Descriptive statistics were used to describe the means and standard deviations of individual items, and to identify floor or ceiling effects. Internal consistency for the seven domains was evaluated using a Cronbach’s alpha, with a Cronbach’s alpha of at least 0.7 considered acceptable (35). Confirmatory factor analysis We conducted confirmatory factor analysis (CFA) to assess the internal structure of the translated eHLQ. CFA was performed as the eHLQ has a pre-specified factor structure. We evaluated the extent to which the items loaded on the seven hypothesized scales (i.e., the latent factors) based on the seven dimensions that the eHLQ intends to measure. CFA was performed using the R package Lavaan in R version R-3.6.1 (63). We fitted a seven-factor CFA model allowing for correlation between latent factors. The Diagonally Weighted Least Squares estimator was used, which is the recommended estimation for ordinal data (64). The CFA provided the standardized and unstandardised factor loadings between item responses and the underlying latent variables. In line with the original eHLQ development study, we report on the robust indexes Comparative Fit Index (CFI), Tucker-Lewis index (TLI), Standard Root Mean Square Residual (SRMR) and Root Mean Square Measure of Approximation (RMSEA). We used the following threshold values for the test of good model fit; CFI >0.95, TLI >0.95, SRMR<0.08 and RMSEA <0.06 and the thresholds: CFI >0.90, TLI >0.90, and RMSEA< 0.08 as indicators for reasonable fit (65). We deemed an item factor loading of 0.4 substantial (66). Items that performed poorly on the above criteria were flagged. To further examine the flagged items, possible model improvements were performed. Invariance testing and multi group comparison Based on previous studies on health literacy and eHealth literacy, and the cognitive interviews, we hypothesized that the demographic characteristics age, gender, educational background and self-reported current diagnosis may affect how the items are interpreted and thus introduce measurement invariance. Hence, we defined the following subgroups prior to performing CFA: Age ≤45 years versus >45 years (median split), gender, high vs. middle and low educational background, and diagnosis yes vs. no. Of the 1650 participants, 467 (28.3%) were male, 1307 (79.2%) were highly educated, and 595 (35.9%) reported to currently have a diagnosis. To investigate measurement invariance within these subgroups we performed invariance testing.
RkJQdWJsaXNoZXIy MTk4NDMw