Concurrent validation of substance abusers self-reports against collateral information: percentage agreement vs. kappa vs. Yule's Y.

The ability for chemical users to give an accurate self-report of substance use vs. abstinence has been questioned. This study investigated its concurrent validity, against collateral ratings. The results indicated that validity of reports of chemical use must be evaluated in the context of the validity of other types of information. Chemical use items were corroborated about as often as such nonchemical use items as reports of emergency room visits, arrests, and hospitalizations, thus arguing against the presence of a specific denial syndrome or overarching tendency toward self-misrepresentation. Relative concurrent validities seemed more a function of such factors as item salience and specificity. No consistent trend in patient over- or underreporting of chemical use was found. The choice of concurrent validation statistic is important and can influence interpretation of results. Current standards such as percentage agreement and kappa were seen as flawed; comparisons of results based on these two measures, as well as Yule's Y led to the conclusion that Yule's Y is the statistic of choice.