Inconsistencies in Adolescent Self-Reported Sexual Behavior: Experience from Four Randomized Controlled Trials

To examine the degree, correlates, and implications of inconsistent self-report data on sexual risk behaviors of adolescents. We analyzed data from four longitudinal group-randomized controlled trials of evidence-based HIV/STI/pregnancy prevention programs in Texas and California from 2000 to 2010. Across- and within-time logical inconsistencies in sexual behavior survey responses were analyzed using multilevel logistic regression. Rates of any inconsistencies ranged from 12 to 18% across the four trials. In all trials, rates were higher in males than in females. Age, normative beliefs, and race/ethnicity were most strongly associated with inconsistencies. We found substantial rates of inconsistencies in adolescents' self-reports of their sexual behavior, which did not occur at random. Studies should routinely report observed rates of inconsistencies and methods used to adjust for them so that any biases in the population to which the study generalizes are understood by public health practitioners and policy-makers looking to adopt programs for their particular population.