Aims This study examines the predictive validity of sensation seeking as a predictor of adolescent substance use, in order to optimize targeting for substance use prevention programs.
Design Longitudinal study.
Setting Random-digit dial telephone survey.
Participants A total of 6522 US adolescents aged 10–14 years at baseline, resurveyed at 8-month intervals for three subsequent waves.
Measurements Two outcomes were assessed—onset of binge drinking (more than five drinks in a short time) and established smoking (>100 cigarettes life-time). Sensation seeking level was assessed at baseline. Logistic regression was used to predict onset of substance use at any follow-up wave as a function of sensation seeking. The receiver operating characteristics curve was used to illustrate how well sensation seeking predicted substance use as a function of different cut-off points for defining high sensation seeking, and area under the receiver operating characteristics curve (AROC) was the metric of predictive validity.
Findings Of 5834 participants with one or more follow-up assessments, 5634 reported no binge drinking and 5802 were not established smokers at baseline, of whom 717 (12.7% of 5634) reported binge drinking and 144 (2.5% of 5802) reported established smoking at one or more follow-up interviews. Sensation seeking predicted binge drinking moderately well [AROC = 0.71 (95% confidence interval 0.69, 0.73)] and was a significantly better predictor of established smoking onset [AROC = 0.80 (0.76, 0.83)]. For binge drinking, predictive validity was significantly lower in blacks; for established smoking it was significantly higher for Hispanics. Implications for two targeting interventions are discussed.
Conclusions Sensation seeking works moderately well at identifying adolescents at risk for onset of binge drinking and established smoking. This study offers a guide for determining the appropriate targeting cut-off value, based on intervention efficacy, costs and risks.