This is an adaptation of a paper I wrote for my MSc course in experimental methods. I've never run the study that I describe - but I would very much like to!
Background
The arrival of the commercially available Internet in the early 1990s, resulted in a paradigmatic shift in healthcare. The provision of resources moved away fromState-led curative care, to digitally-enabled individual-as-expert preventative care . In this ‘info-liberal’ paradigm individuals allegedly become empowered through Garcia's (2014) micro-cycle of self-reflection in which they gain knowledge and awareness and take proactive action based on this knowledge, after a period of self-reflection.
Justification for this shift rests on the assumption that access to health information automatically translates into rational and reasoned action.However, the causal chain in this argument is missing. It is not clear:
How access to online health information translates into knowledge and action;
How the heterogeneity of online channels and populations affect this process; nor
How the actions triggered by information consumption are constrained by gender, race , income and education levels, confidence in digital and health literacy skills, or a combination of all of these factors (Bravo et al 2015).
A 2018 study sought to address some of these knowledge gaps by investigating how health literacy, defined by the Institute of Medicine as the ability to “obtain, process, and understand basic health information and services needed to make appropriate health decisions,”affects which sources of health information consumers trust. 618 members of GfK’s Knowledge Panel were recruited to complete a survey which tested their health literacy level, asked them to identify which of 25 sources of health information they used, and how much they trusted each source. Multiple regression analysis found that that people with lower health literacy were less likely to trust medical websites and more likely to trust social media, blogs and celebrity webpages .
These findings serve to highlight the fact that the relationship between online health information consumption and proactive health action is unlikely to be as straightforward as the information = empowerment argument implies. However, the cross-sectional design of the study limits our ability to draw causal inferences from it. As such it remains unclear why those with low health literacy trust social media more than medical websites as sources of health information, making the knowledge harder to act on. This raises the question: "why do individuals with low eHealth literacy trust social media as sources of health information more than medical websites?"
The Theory
Trust in a source of health information is an affect produced at least partly by perceived credibility. In the offline world, this is largely controlled by the gatekeeping function performed by clinicians. In an online world, however, this gatekeeping function is removed. This means that a far greater burden is placed on individual Internet users to make the necessary judgements about credibility and to ultimately determine which sources they trust.
According to sense-making theory the ways in which individuals make such subjective decisions about the overall credibility and usability of information, varies drastically depending on their individual context and circumstances. In the search for online health information, one of the main contextual factors influencing the decision-making process, is relevant literacy level. Health literacy has been measured in more than 51 different ways many of which require face-to-face assessment. This suggests limited construct validity and applicability to an online context. In contrast, eHealth literacy, which was conceptualised in 2006, is based on social cognitive and self-efficacy theories and, has one primary validated measurement scale: eHEALS.
As measured by eHEALS, eHealth literacy is composed of six core literacy skills which combine to provide individuals with the ability to work with technology, critically assess quality and credibility of information, and navigate to the sources that provide them with the specific information they need to make a decision. It has been demonstrated that those with high eHealth literacy are able to successfully evaluate the credibility and relevance of online health information following great scrutiny, whereas those with low eHealth literacy often struggle to locate and understand online health information. This difficulty lowers their self-efficacy, distorts their perception of source credibility and negatively impacts perceived trustworthiness of the source.
This creates a need for individuals with low eHealth literacy to find an alternative means of determining trustworthiness in online sources of information, such as social endorsement. Visible social endorsement, such as ‘likes,’ enables those with low eHealth literacy to determine trust based on the bandwagon heuristic and assume that if the source has already been deemed valid by others, then it safe for them to trust it too. Traditional medical websites afford those with low eHealth literacy no such alternative means of determining credibility and trust. This suggests a causal chain that can be tested: social endorsement determines the perceived trustworthiness of an online source of health information, and the extent to which this social endorsement is relied upon (i.e. the strength of the relationship) varies according to eHealth Literacy. Or, in statistical terms, the effect of social endorsement on perceived source trustworthiness is moderated by eHealth Literacy. This can be translated into the following hypotheses:
H0 (null): eHealth literacy has no effect on the relationship between social endorsement and perceived trustworthiness of online health information sources
H1: Increased levels of social endorsement will result in increased perceived trust in online sources of health information
H2: Higher eHealth literacy will result in higher levels of trust in online sources of health information (Paige et al 2018)
H3: Increasing eHealth literacy will decrease the effect of increased social endorsement on increased trust (i.e. the moderator variable, eHealth Literacy, will have a buffering effect
Testing the hypotheses
One way to test these hypotheses would be through an 'experiment.' Specifically, I propose that you could use a a 2 (social media x medical website) by 3 (no social endorsement, low social endorsement, high social endorsement) within-participants experimental design (Smith 2014).
First, participants would take a pre-test questionnaire, asking:
Basic demographic questions
Their current health status (whether or not they are currently living with a long-term health condition)
The extent to which they agree or disagree (7-point Likert scale) with a series of statements to determine their level of health consciousness
The extent to which they agree or disagree (5-point Likert scale) with a series of statements to determine their eHealth literacy
The first three sets of questions are included because previous studies have identified that these are all potentially confounding variables in the relationship between eHealth literacy and assessment of health sources and will therefore need to be controlled for during the analysis stage. For the purposes of maintaining construct validity the questions used to measure health consciousness and eHealth literacy level would both be taken from validated scales from Ophius (1989) and Chen (2011); and from Norman and Skinner (2006) respectively.
Following completion of the pre-test questionnaire, all participants would be shown a screenshot of a webpage from WebMD (a traditional medical website with no social endorsement) providing advice on daily calorie intake. They would then be asked the question from the original study: “how much do you trust the information from the previous source?” (not at all, some, quite a bit, a great deal). This would be repeated with each participant randomly seeing either the a or b option for each of the three sources below:
A Facebook post providing advice on how to reduce bloating either:
High social endorsement: 150 likes/emoji reactions, 35 comments, 50 shares
Low social endorsement: 3 likes/emoji reactions, 0 comments, 2 shares
A Tweet proclaiming the benefits of the DASH and MIND diets either:
High social endorsement: 30 retweets, 150 likes
Low social endorsement: 2 retweets, 3 likes
A Medium blogpost discussing the 5 most basic rules for health and fitness either:
High social endorsement: 150 ‘claps’
Low social endorsement: 2 ‘claps’
The levels of high and low social endorsement, as measured in virality metrics, that I suggest above are based on those used by Kim (2018). Each stimulus should be based on an edited version of a real-post, and participants should be told that the identity of each post has been protected for privacy reasons. In addition, all should provide preventative health advice, as it's been shown that different threat levels can alter how people make decisions about trustworthiness, and thus the perceived level of threat needs to remain static. Similarly, I suggest WebMD be used rather than NHS.uk as the content on NHS.uk is specifically designed to meet eHealth literacy levels and public trust in the NHS brand is so high this could be confounding.
Analysis
So how would the results from this proposed experiment be used to answer the research question and test the hypothesis? By using multiple regression of course!
More specifically, as is standard in moderation analyses, I would propose that moderated hierarchical multiple regression be used to estimate whether the prediction of trust in source (Y), from social endorsement (X) differs across levels of eHealth literacy (Z) . This would involve including the moderator variable (eHealth literacy) as an interaction term in the regression equation: Y = i5 + β1X + β2Z + β3XZ + e5 where β1 is the coefficient relating social endorsement (none, high or low) to trust (low, moderate or high) (H1). β2 is the coefficient relating eHealth literacy (low or high) to trust (H2), and β3 is the interaction coefficient providing an estimate of the interaction effect of social endorsement and eHealth Literacy on predicting Y (H3) (i5 is the intercept in the equation, and e5 is the residual. The control variables should be treated in keeping with standard practice of hierarchical multiple regression and all variables should be standardized to avoid multicollinearity.
For the null hypothesis to be rejected the effects of both β1 and β2 would have to be significant and the effect of β3 should be both significant in and of itself as well as significantly different to β2.
If the null hypothesis is rejected and H3 is supported, this finding would have a number of implications for online health promotion strategies and the “information = empowerment” narrative because, as fear-based misinformation about health is often the more ‘engaging’ (liked) content on social media than factual medical advice, those with eHealth literacy could be being ‘empowered’ by false information. This puts pressure on public health promoters to decide when to intervene and how to more effectively target accurate campaigns at those with low eHealth literacy.
If the null hypothesis is not rejected, this could be due to insufficient power, or it could mean that sample of participants was biased towards those with high eHealth literacy making it difficult to detect the significance of the moderation effect. Alternatively, it could mean that there are issues with reliability or that one (or more) of the myriad features of social media platforms is acting as an additional moderator and/or mediator. Given these potential issues, and the fact that whilst this study would have high internal validity, but limited external validity and consequential generalizability, this investigation would have to be seen as part of an ongoing research program rather than an isolated investigation.
Comments