RE: RELIABILITY OF SELF-RATED HEALTH IN US ADULTS

In their analysis of the reliability of self-rated health among US adults (1), Drs. Zajacova and Dowd concluded that, ‘‘1) [T]here is a substantial amount of error in individuals’ self-assessment of health and 2) reliability is worse for disadvantaged sociodemographic groups, potentially biasing estimates of health inequalities’’ (1, p. 977). Although their study marks a relevant contribution to the understanding of this important and versatile variable, in my opinion, the use of the terms ‘‘error’’ and ‘‘reliability’’ might induce misunderstanding of some key aspects, mainly because both are concepts in the realm of objectiveness, whereas self-rated health is an inherently subjective variable that is more akin to perception than to systematic, objective evaluation. Although error, a concept more closely related to the notion of validity, implies disparity vis-à-vis a referent, there was no such referent in the study in question, no gold standard, inasmuch as it is not clear whether the benchmark was the first assessment, which was conducted in a home setting, or the second, which was conducted 30 days later in a medical setting. In general, I would not describe any particular subject’s self-assessment as an error, provided that the questions were reasonably understood (e.g., excluding people with cognitive impairment). By the same token, it would be advisable to avoid the term reliability because it conveys a notion of an objective measurement process of some underlying stable parameter that could eventually be determined. It would therefore not seem to be appropriate to talk of reliability in the case of subjective assessments in which change is arguably a more natural state than is constancy; any change in a health rating must necessarily be attributable to a change in its determinants, and yet objective health aspects comprise only part of such determinants. It is true to say that using the term reliability might be more permissible if self-rated health were explicitly used only as a summary indicator of some objective health state because in a situation of that type, an underlying referent, albeit implied, would be assumed to exist. In this regard, it would be relevant to ascertain about what people believed they were being questioned. Some persons might possibly understand that they were expected to make an assessment based on objective medical facts, but even those people generally cannot avoid adding some nonnegligible subjective weight to their responses. To sum up, the term error should not be used, and one would also do well to avoid the term reliability. Perhaps the use of closely related terms, such as ‘‘variability’’ or ‘‘degree of stability,’’ might better convey the essence of the phenomenon. In light of the above, the wording of the original conclusion would be amended to read, ‘‘There is a substantial amount of short-term variability in individuals’ self-assessment of health, especially for disadvantaged sociodemographic groups.’’