Inaccuracies in self-report measures of adolescent risk-taking behaviors

Education, Training, & Research Assocs, Scotts Valley CA

Investigators

Linked publications, trials & patents

Abstract

DESCRIPTION (provided by applicant): The proposed R03 research project features secondary analyses of datasets from four longitudinal group- randomized trials of HIV/STI/pregnancy prevention programs to provide the field of sexual and reproductive health with more detailed information than has previously been available about the degree and nature of inconsistencies in data from self-report measures of high-risk adolescents' sexual behaviors - critical information for improving the accuracy of future measures. Self-report measures are the predominant method for monitoring of sexual risk behaviors and evaluation of the effectiveness of prevention interventions (McFarlane and St. Lawrence, 1999; Upchurch et al., 2002; McAuliffe et al., 2007; Palen et al., 2008; Rose et al., 2009). Policy-makers set priorities based on trends in these data and intervention researchers use them to make conclusions about the effectiveness of prevention programs (Upchurch et al., 2002; Brener et al., 2003; Beguy et al., 2009; Rosenbaum, 2009). Yet studies show rates of inconsistencies in U.S. adolescents' self- reports of their own sexual behaviors may be substantial, ranging from 8% to 15% (Rogers et al., 1982; Alexander, et al., 1993; Upchurch et al., 2002; Brener et al., 2003; Rosenbaum, 2006; Rosenbaum, 2009). Furthermore, results from a few studies examining patterns in inconsistencies across demographic/risk subgroups of adolescents have been mixed, suggesting a need for additional research. The proposed study would examine overall rates of within- and across-time inconsistencies in reports of sexual behaviors (e.g., lifetime sexual experience) for each of the fours studies separately, and investigate patterns in rates across various subgroups of youth defined by demographic characteristics and alcohol/drug use. The study would identify factors significantly associated with inconsistent reporting (e.g., gender, alcohol/drug use) using multilevel logistic regression models. The study also would conduct post-hoc power analyses to provide new information on the impact of removing different types of inconsistent cases on power; it would examine impact on generalizability by determining how data would need to be re-weighted if inconsistent cases were removed. Finally, results of the study would move the field forward in improving the accuracy of measurement tools by identifying groups of youth with whom to explore, in a future study (e.g., R21 measurement development), potential reasons behind inconsistencies in adolescents' self-reported sexual behaviors. Empirical information about the reasons behind inconsistencies is essential for guiding strategies to improve the accuracy of self- report measurement tools of adolescent sexual behaviors.