Time-Varying Predictors of Sexual Behaviors in Young Adult Women

Background: Sexual behavior carries different meanings and risks across different periods of the lifespan and in different contexts.  For example, sexual behavior may be associated with problem behaviors and poorer mental health in early adolescence, but improved mental health and relationship status in adulthood.  Similarly, couples may use condoms consistently at the beginning, but not later, in a sexual partnership. Thus, it is important to use data collection procedures and statistical methods that can capture the dynamic processes involved in sexual behavior.  In this paper, we demonstrate how a new analytic method, the time-varying effect model (TVEM), can be applied to understand how sexual behavior unfolds in young women at different ages and in different stages of sexual partnerships.

Method. Data are from a longitudinal study of women who were recruited from public records when they were 18-20 years old and provided weekly reports of their sexual behaviors and romantic relationships for up to 2.5 years (N=1,003, 35% African American, Mage=18.7). Outcome measures include engaging in sexual behavior and sex without a condom in the past week, and predictors include attitudes about sex and contraception, peer norms, religiosity, and partner characteristics.  Data are analyzed using logistic TVEMs, which are flexible, nonparametric models for estimating associations between variables over time.

Results. TVEMs demonstrate the ways in which predictors of sexual behaviors vary over time, including both developmental time (i.e. age) and time in a specific partnership. As coefficients are estimated for points in near continuous time, results will be presented primarily as figures showing changes in regression coefficients over time. Preliminary results show that the impact of religiosity on odds of having sex changes between ages 18 and 23.  At age 18, women who are more religious are less likely to have sex.  This effect dissipates with age, such that there is no significant association between ages 19 to 21.  However, by age 21, individuals who are more religious are more likely to have had sex in a given week.

Conclusions. Results demonstrate the importance of examining how predictors of sexual behaviors change over time and across different stages of relationships.  Discussion will focus on the meaning of specific findings, as well as general issues related to data collection, spacing of measurement occasions, and statistical analysis, and ways in which collection of long-term intensive longitudinal data and TVEM can be applied to understand and prevent sexual risk behavior in future studies.