Hello,
Apologizes for cross-posting.
There has been a growing trend in using assessments of absenteeism and poor performance (presenteeism) that build in attributions for the behavior. For example,....
In the past year, how many days of work have you missed due to personal illness?
In the past 2 weeks, how much of the time did your physical health or emotional problems make it difficult for you to do the following?'' (followed by a list of items).
Used in purely descriptive terms, answers to these types of questions seem potentially interesting. However, they aren't assessing a construct per se, they are assessing an implied or inferred association. These measures become a problems when they are correlated with variables that are similar to the attributions. If you want to know if health in general or some particular health problem is related to reduced performance and absenteeism, these measures would be troublesome in that any association may be inflated because the antecedent and outcome are both measures of "illness." Even if it an antecedent was something else, say exposure to harassment, does a positive correlation mean harassment is associated with illness, absenteeism, or both? What if one wanted to test whether poor health mediates the association of harassment to performance or absenteeism? Using these outcomes would seem problematic.
Some researchers, and some measures are designed to, insert specific illnesses or conditions rather than something broad like "illness."
Here is the most egregious example I've found to estimate the association between alcohol use and absenteeism:
First, measure "alcohol-related" absenteeism:
How many days have you been absent because of your drinking during the past 12 months?
Then regress this absenteeism variable on some covariates and a measure of alcohol use:
How often did you drink during the past 12 months.
It might not be surprising that this "published" association is too large to be believable as an estimate of a putative causal effect of alcohol on absenteeism. It is essentially an assessment of convergent validity--do two measures of alcohol use correlate. Perhaps more amazing is this example was published in a respectable journal.
Clearly, not all associations based on attribution-laden measures of absenteeism and presenteeism are this problematic, but they all seem to open the door to confounding and ambiguous interpretation, and spurious and inflated associations.
So, has anyone seen an article that discusses this issue? I can't find one.
Thanks,
Mike Frone