Test-retest reliability of adolescents' self-report of substance use

Background: Adolescent substance abuse is a serious problem for which effective interventions are needed. To conduct trials of new therapies, investigators need reliable means of identifying potential participants and of measuring outcomes. The objective of this study was to determine the 1-week test-retest reliability of the CRAFFT screening test and of the timeline follow-back (TLFB) calendar method for measuring alcohol, cannabis, and other drug use. Methods: Ninety-three 12- to 18-year-old patients presenting for routine medical care to three urban adolescent clinics were administered the CRAFFT screen in both lifetime and past-year versions and a 90-day TLFB. Both measures were completed on the day of the clinic visit and again 1 week later. We computed K coefficients and the intraclass correlation coefficient (ICC) for the CRAFFT and computed the ICC separately for TLFB self-reports of alcohol and cannabis. Results: For the CRAFFT, kappa for individual items ranged from 0.31 to 0.86, and the ICC was 0.93 (95% confidence interval, 0.90-0.95). However, the total score of the lifetime CRAFFT at time 2 was significantly lower than at time 1; there was no difference in time 1 and time 2 total scores for the past-year version. The ICCs for past-90-days TLFB variables were as follows: drinking days, 0.92; drinks per occasion, 0.87; cannabis days, 0.83; and joints per occasion, 0.76. Past-30-day and past-60-day intervals compared favorably to past-90-day intervals. Conclusions: The CRAFFT screen is a reliable means of screening adolescents for substance abuse, although we recommend using the past-year version. The TLFB is a reliable method of quantifying adolescents' alcohol and cannabis use at intervals of 30, 60, or 90 days.