Sexual Concurrence Patterns in the U.S.

Battelle Centers/Pub Hlth Res &Evaluatn, Columbus OH

Investigators

Abstract

DESCRIPTION (provided by applicant): The overarching objective of the proposed research is a multi-level examination of sexual concurrency and mixing patterns among U.S. men and women. Specifically, we will examine: (1) the patterns of concurrent partnerships, and gender differences in these patterns as a function of individual, relationship, and community characteristics; (2) the individual, relationship, and community characteristics associated with the risk of entry and exit from concurrent partnerships; and gender differences in the effects of these characteristics on that risk; (3) the differences between continually monogamous and initial concurrent partnerships and between initial concurrent and secondary concurrent partnerships with respect to partner characteristics (i.e., mixing patterns) and HIV-risk sexual behaviors, including alcohol and drug use prior to and with sex. We have two corollary aims. These are, to examine: (4) the cross-sectional change in the patterns of concurrent partnerships over time, and the change in the effects of individual, relationship, and community characteristics on the likelihood of formation and dissolution of concurrent partnerships; and (5) the cross-sectional change, overtime, in HIV-risk sexual behaviors, including alcohol and drug use, within different types of partnerships, and the change in the effects of these characteristics on HIV-risk sexual behaviors. We will use data from four national surveys of men and women: The 2002 National Survey of Family Growth (NSFG), the 1996 National Sexual Health Survey (NSHS), and the 1991 National Survey of Men (NSM) and the National Survey of Women (NSW). The research is guided by the expectancy-value theory, and the analyses are based on a modified version of the Subjective Expected Utility model, hi the analysis of the risk of entry into and exit from concurrent partnerships we will use hazard models. Binary outcome measures will be analyzed using logistic regression. The analysis of risk perceptions which are ordered scales will use ordered logit models. When the outcomes measures are nominal unordered categories we will use multinomial logit models, and when the outcome measures are skewed toward zero we will use tobit models, which can incorporate information on the proportion of the sample with outcomes of zero.