An empirical comparison of three psychometric measures of adolescent substance use (2015)

Undergraduate: Miriam Brinberg

Faculty Advisor: Patrick Curran
Department: Psychology & Neuroscience

---

Adolescent substance use is common and often accompanied by many negative consequences. The psychometric assessment of substance use, however, is highly varied and may misrepresent actual substance use. This study examined three competing measurement models of substance use and their differential predictability of academic achievement and internalizing and externalizing symptomatology. Participants were drawn from the public-use Add Health data set (N = 5,857). The three measurement models were determined from seven items that assessed the lifetime use of different drugs. The first model grouped participants as abstainers and users. The second model refined substance users by accounting for the proportion of the seven drugs they endorsed. The final model created factor scores through confirmatory factor analyses that allows each item to be differentially weighted as a function of severity. Nine separate regression analyses were estimated in which age, race, gender, and measurement model predicted either academic achievement or externalizing or internalizing symptomatology. These analyses indicated that all three substance use measures were significant predictors of each outcome measure, but the proportion and factor scores models accounted for substantially more of the variance within each model. These results suggest the proportion or factor scores would be better predictors of substance use, although the choice of model should be based on relevant theory.