Maricianah,
Excellent and very interesting examples. I think what you are suggesting for validation of the health literacy measure is an assessment of construct validity - does it correlate with what we expect it to? This is extremely useful.
For the disparities paper, note that regression models can be used to derive either absolute or relative measures of disparities. Specifically, with binary outcomes, we usually use a logit model (as they did), which directly provides odds ratios. These odds ratios are a relative measure, and that's what they report in their abstract.
However, it is also possible to estimate a linear regression model with a binary outcome (even though everybody trained in epidemiology will tut tut disapprove of you for this scandalous behavior because the predictions will often fall out of range of the binary measure). If you endure the disciplinary disapproval though, the effect estimate from such a linear model is actually the risk difference: the difference in the probability of the outcome between the reference category and the group to which the estimate refers.
But it doesn't seem like that's what they did in this paper - they used logit models so derived a relative effect measure. Epidemiologists go to logistic regression very quickly but note that's somewhat of a disciplinary quirk. Economists are more likely to use linear models, even with a binary outcome, or perhaps a probit model.
Maria