Almost all young adults in the UK are shown to be recent internet users [1] with teenagers present online for over 6 hours a day [2]. Indeed, most young adults consider the internet a valid source of health information [3,4] and compared with “digital immigrants,” “digital natives” (ie, those who have grown up in the digital age) seek out more health information electronically [5]. While the internet has transformed and widened access to health information, there has been concern about the quality of electronic health (eHealth) resources [6,7] and indeed whether people can select good quality health information on the internet. Undoubtedly, increasing the health literacy of populations is now seen as essential and is fast becoming a global challenge [8]. In England, 15% of adults are classed as functionally illiterate [9] and 32 million adults in the United States are unable to read [10]. However, successfully accessing eHealth resources goes beyond functional literacy skills and requires “eHealth literacy,” defined as the ability to “seek out, find, evaluate and appraise, integrate, and apply what is gained in electronic environments toward solving a health problem” ([11], p. 2).

Intuitively one would assume that young people as “digital natives” would be competent users of the technology they have been brought up with; however, this is not always the case [12]. Equally they are not always capable users of eHealth resources [13]. This lack of competency and fundamental concerns around the extent of misinformation in eHealth resources [14] have the potential to impact public health in terms of individual health outcomes and unnecessary burdens on health care systems [15]. Therefore, it is important to know whether young people can discern if the information gained from eHealth resources is of good quality and appropriate to their needs. To this end, Norman and Skinner [16] have developed a measure of perceived eHealth literacy—the “eHealth Literacy Scale” (eHEALS). Good reliability and validity from the eHEALS are demonstrated in young people in the United States [17] and the measure has been successfully translated into several languages [18-20]. Thus, if we are to successfully measure eHealth literacy, it is also important to understand factors that may influence the perceived ability of teenagers and young adults (TYAs) to successfully navigate internet-provided health resources. This is particularly crucial as this is the first generation to have grown up with access to eHealth resources and will continue to “Google it” when faced with a health problem throughout their lives. Therefore, it is important to gain a deeper understanding of what factors potentially influence eHealth literacy in young people. Norman and Skinner [11] propose that eHealth literacy is underpinned by a set of complex skills depicted by the 6-factor Lily model. The model proposes the following 6 overlapping literacies that determine and influence eHealth literacy: traditional literacy, health literacy, information literacy, scientific literacy, media literacy, and computer literacy. Here we explore putative variables which could enable the exploration of eHealth literacy and the 6-factor Lily model in young adults today.

The primary aim was to explore whether factors potentially underpinning the 6 facets of the Lily model [11] (including NVS [FHL], IHBs, NFC, and GSE) are associated with and are significant predictors of eHealth literacy in young people measured by the eHEALS [16].

A secondary aim was to explore the underlying constructs of the eHEALS using an exploratory factor analysis (EFA) in a sample of university students in the UK.

Participants (n=188) were undergraduate psychology students from a UK University recruited via opportunity sampling in research methods sessions. Of these, 167 participants were female (88.8%), with a mean age of 20.13 (SD 2.16) years. The majority of participants identified their ethnicity as White British 159/188 (84.6%) and said that English was the main language spoken at home (181/188, 96.3%). Other participants described their ethnicity as Mixed 4/188 (2.1%); Pakistani, Indian, or Asian 13/188 (6.9%); any other White background 2/188 (1.1%); or Black 2/188 (1.1%). The majority were unmarried 182/188 (96.8%) and living in shared accommodation 122/188 (64.9%). Most participants 139/188 (73.9%) had not seen a health professional in the past 2 weeks. Of those that had, most had seen a general practitioner (family doctor; n=26) or a physiotherapist (n=4).

Ethical approval from the Leeds Beckett Psychology Ethics Committee was granted on December 12, 2017 (number: PH/AW/121217). Participants took part in this study in order to generate data for them to analyze in a research methods module, and gave informed consent for these data to be used in further research. All measures were programmed into Qualtrics, a web-based survey tool, and participants completed these in-exam conditions during research methods sessions over the course of 1 week in March 2018.

All data analyses were conducted using IBM SPSS Statistics (version 26). Internal consistency of the scales was examined using the Cronbach alpha coefficient. The psychometric properties of the eHEALS in a UK sample were explored using EFA. We initially examined correlations between all key variables. We had theoretical grounds to suggest that GSE would be a significant predictor of eHealth literacy, therefore a hierarchical multiple regression was performed by entering GSE first and then examining the impact of NFC, IHBs, NVS, and GSE on the outcome variable eHEALS. For the NVS, 18% of the data were missing, therefore mean scores were inputted; missing data were <0.01% for other variables.

Data assumption tests were performed prior to conducting the linear regression, including Cook distance, collinearity, variance inflation factor, Durbin–Watson, and homoscedasticity. To test for normality, skewness and kurtosis values were computed prior to correlation analysis.

An EFA was used to examine the eHEALS structural validity on the 8 data item set (N=188 cases). The minimum recommended sample size to conduct an EFA is 100 [47] and if factors emerge with 4 or more loadings over 0.6, then this would be deemed reliable regardless of sample size [48]. In addition to the total variance explained, the scree plot, eigenvalues, and component loadings were assessed to verify the factor structure of the eHEALS in this population.

In this exploratory study we have been the first to investigate relationships and potential predictors of key variables influencing eHealth literacy sample of UK University students. A highly significant positive relationship was found between eHEALS and GSE scores, but eHealth literacy did not significantly correlate with any other factor. However, significant positive correlations between NFC and the NVS, and between NFC and GSE were demonstrated as well as a significant negative association between NFC and IHB. Our secondary aim was to explore the underlying construct of the eHEALS in a UK TYA sample using EFA, where we found 2 underlying factors within the scale related to information acquisition and information application.

Despite the expected negatively skewed distribution indicating higher scores on the NVS in our student sample, FHL measured by the NVS was not found to be a significant predictor of eHealth literacy. Previously Del Giudice and colleagues [21] demonstrated large associations between functional and eHealth literacy; however, in their study, FHL was not directly measured but rather assumed (as a proxy measure) based on studying or working in the health sector (eg, physicians, nurses, and allied health professions). Further, Del Giudice et al [21] found that for older and better educated participants eHEALS scores were higher. Our sample comprised psychology students (a health-focused subject), but perhaps the younger age of our sample (mean age 20) could have been an influencing factor in the lack of significant association with the eHEALS scores.

In line with the literature, those who scored highly in our sample on the NVS (a test of functional literacy) also scored highly on the NFC scale. This suggests an overlap between these 2 variables in that someone who prefers effortful cognitive activity would be more likely to score highly on a functional literacy test [26]. Our findings do not support the health literacy aspect of the Lily model in terms of demonstrating a relationship between functional and eHealth literacy. However, the expected negatively skewed scores in this educated sample and the internal consistency of .60 of this scale would suggest proceeding with caution in interpretation of our findings. In future studies it may be useful to explore FHL with the NVS in a more diverse population of young people who are either unemployed or who have not had a university education, as there is an urgent need to explore the factors driving FHL in all young people [50] to enable successful health promotion strategies.

The significant negative association between NFC and IHBs was as predicted, as IHBs are usually strongly held and people displaying IHBs have a limited need or desire to acquire new information to challenge or inform these beliefs [32]. Although counterintuitively, this did not transfer into a tangible negative association with eHealth literacy.

SE has been shown to be a pivotal component in the adoption and execution of healthy behaviors [51], and that it being a significant predictor of eHealth literacy would seem entirely plausible. Previously, internet SE was deemed to play a key role in the process of using the internet to acquire health information [52]. With this in mind it may be useful to explore the relationship between eHealth literacy and a measure of health-related SE (eg, [53]). Using a dedicated health SE measure may have accounted for more variance in the model and therefore may be a greater predictor. Interestingly, others have argued that the eHEALS actually measures SE rather than eHealth literacy [29]; indeed, the question prefixes (eg, “I know how to” and “I have the skills to”) would seem to tap into confidence and SE rather than assess performance. Perhaps, implementation of internet skills could not be predicted from high eHEALS scores, suggesting an incongruence between perceived and actual eHealth literacy [54]. We echo the thoughts of Britt and Hatten [29] who call for the relationship between SE and eHealth literacy to be fully explored.

We found a lack of significant correlation and predictive impact of NFC on the outcome variable (eHEALS) in contrast to the US study of young university students in which an association between the 2 variables was demonstrated [29]. It is useful to try to explain the reasons for failure to replicate these findings in a UK sample. The mean eHEALS scores and standard deviations were very similar in this study and the one conducted in the United States [29], and both samples were of a similar age. However, the US study had a larger sample yielding greater power to detect significant associations. Moreover, our UK sample produced higher NFC scores than participants in the US study, but the latter did not provide information on the types of courses the students were enrolled in. Conceivably, the type of course influenced the results, as Del Giudice and colleagues [21] found that participants’ exposure to health-related study was associated with higher eHealth literacy scores.

Previously the eHEALS was thought to be a unidimensional scale [16,17]; however, in the UK a confirmatory factor analysis and structured equation modeling in a large study with baby boomers in the United States, UK, and New Zealand (born between 1946 and 1954) found the eHEALS to have 3 distinct scales related to the Lily model [43]. They found items 1 and 2 related to awareness, items 3-5 related to skills to access resources, and items 6-8 evaluation of resources (self-efficacy). The authors recognize that more work needs to be done, particularly in other age groups to verify the 3-factor structure. Our findings that the eHEALS related to 2 distinct constructs, namely, information acquisition (items 1-3) and information application (items 4-8), concur with those of Soellner et al [55], who found the 2-factor structure to be more compelling than a single one. However, we would recommend continuing to explore the construct validity of the scale in different populations.

This was a cross-sectional design and as such only measured eHealth literacy at one time point, although it may be more revealing to measure this longitudinally to explore temporal changes and responses to changing circumstance (eg, an individual’s own health concerns). The skewed distribution of the NVS data and low internal consistency ensure that we proceed with caution and thus cannot conclusively state that FHL does not influence eHealth literacy in a UK TYA population. However, we also must question “Does perceived eHealth literacy (as measured by the eHEALS) translate into competent performance on the internet to gain health information?” Others have also found that the expected strong positive relationships between eHEALS scores and internet performance were absent [53]. This suggests that future studies should also focus on measuring practical internet tasks along with perceived eHealth literacy as a comprehensive measurement of true eHealth literacy.

In this study, because GSE only accounted for a small proportion of variance in our predictive model, we must speculate that alternative, more potent predictors of eHealth literacy exist. Gilstad [56] has proposed further aspects to eHealth literacy, thereby expanding the Lily model proposed by Norman and Skinner [11]. Four additional aspects to the Lily model were proposed: bodily experience, procedural literacy, contextual and cultural literacy, and communicative expertise. These additional factors are particularly pertinent to young people as they navigate seeking health information online, particularly knowledge of the norms, values, rules, and regulations in social situations (contextual and cultural literacy) and bodily experience (the ability to identify a health problem in one’s own body). A recent work [57] demonstrated that delayed help seeking for potential breast cancer in females was associated with dissatisfaction with their breasts. Given that body dissatisfaction increases for both males and females as they transition to young adulthood [58], it would be interesting to explore whether body dissatisfaction could impact on their eHealth literacy.

We are reassured that NHS England [59] is working toward improving health literacy to reduce health inequalities, and perhaps there should be a greater directive to expand this initiative to include eHealth literacy. eHealth literacy should not be seen as a static state, but rather as a dynamic evolving skill set that will develop over time and in response to individual need and circumstance [11]. This would seem a logical step forward to serve digitally native young people as these are the generation that will “Google it” for health information into adulthood.

In conclusion, this is the first study in the UK to explore relationships between these key variables in a TYA population, and to perform EFA on the eHEALS with a TYA sample in the UK. As such, we can confirm the stability of the scale. The findings that SE has a major influence on eHealth literacy should consolidate its status as underpinning the Lily model, and as a fundamental starting point from where eHealth literacy is developed and understood. Future studies will explore physical and mental health status, health SE, and body image as potential predictors of eHealth literacy in more diverse TYA populations.