Using sensation seeking to target adolescents for substance use interventions

Abstract

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.

Keywords: Smoking; adolescent; binge drinking; false positive rate; longitudinal; predictive validity; receiver operating characteristic curve; sensation seeking; sensitivity.

Figure 1

ROC curve for sensation seeking as a predictor of binge drinking, based on a bivariate logistic regression (n = 5634, likelihood ratio chi square with 1 degree for freedom = 336.12). Note: The numbers above the ROC curve reflect the percentage of all adolescents identified as at risk for each sensation seeking cut-off. The median split cut-off is 52.

Figure 2

ROC curve for sensation seeking as a predictor of established smoking onset, based on a bivariate logistic regression (n = 5802, likelihood ratio chi square with 1 degree for freedom = 170). Note: The numbers above the ROC curve reflect the percentage of all adolescents identified as at risk for each sensation seeking cut-off. The median split cut-off is 52.