Back to Annual Meeting

Back to Annual Meeting

APHA Scientific Session and Event Listing

Janet E. Rosenbaum, Health Policy PhD program, Statistics Department, Harvard University, 15 Leonard Av, Cambridge, MA 02139, 617-499-4362, jerosenb@fas.harvard.edu

Objective: Surveys are the major source of data for understanding adolescent health risk behaviors. Some adolescents over-report their actual risk behaviors and others under-report. Under- and over-reports can be detected if adolescents give logically inconsistent reports of risk behaviors: if an adolescent reports having engaged in a risk behavior and later reports having never engaged in it, at least one of the two reports must be false. For frequently-recanted risk behaviors, the proportion of adolescents reporting the behavior at a single point in time may not accurately measure the risk behavior's prevalence. Methods: A Markov Chain Monte Carlo method to estimate the prevalence of a disease without a gold standard test (Black & Craig 2002) is modified for adolescent risk behavior. Sensitivity and specificity of self-report for each risk behavior are modeled to depend on age, gender, and other covariates. Results: More than 10% of adolescents recant their earlier reports of sexual intercourse (11%), age of first sex (64%), pregnancy (19%), and virginity pledges (53%). Prevalence estimates from the model differ from standard cross-sectional estimates. Conclusions: Adolescent risk behavior prevalence estimates are more uncertain than is implied by conventional meaures of uncertainty, such as standard errors. Specificity and sensitivity of self-report for risk behaviors can be used in future adolescent health research.

Learning Objectives: At the conclusion of the session, the participant (learner) in this session will be able to

• List the seven self-reported behaviors recanted by more than 10% of respondents in the Add Health survey.
• Describe the Bayesian model for determining the prevalence of a disease with no gold standard test.
• Construct a model to estimate the prevalence of risk behaviors with high recanting.

Keywords: Biostatistics, Adolescent Health

Presenting author's disclosure statement:

Any relevant financial relationships? No