Retrospective use of alcohol and other substances by college students: Psychometric properties of a new measure

Using a new retrospective measure, interviews regarding alcohol and other substance use for the 5-year period preceding entry to college were conducted twice (with a 3-week interval) among 111 college students. Test-retest reliability was uniformly high across alcohol and substance use items (mean of Pearson correlations=.85), as were intraclass correlation coefficients (mean=.81) and kappa coefficients. A latent growth model (LGM) was specified and evaluated to test a hypothesized linear trend in alcohol use across the 5-year interval referenced by the retrospective measure. Furthermore, convergent and discriminant validity were provided as the slope (or rate of acceleration) parameter associated with the retrospective assessment predicted current alcohol disorders but not other disorders (e.g., depressive or anxiety disorders). The intercept predicted both current substance disorders (principally marijuana disorders) and anxiety disorders. Findings were discussed with regard to the use of this retrospective interview measure to screen for serious alcohol and other substance use prior to new students entering college, thereby affording opportunity for early interventions.