Session: A New Approach to Model the Etiology of Health Risk Behavior: Estimating Change over Historical Time, Developmental Time, and Time from an Event (Society for Prevention Research 22nd Annual Meeting)

Health risk behaviors, including substance use and risky sexual behavior, are highly dynamic in terms of both their prevalence across population subgroups and the salience of different risk factors throughout development. A new statistical approach, the time-varying effect model (TVEM), presents new opportunities to reveal nuanced changes over time in health risk behaviors and to provide insights into improved intervention efforts that are tailored with respect to population subgroups and – importantly – to time. This symposium presents three compelling applications of TVEM to very different types of data on health risk behaviors. Each talk presents a unique metric of time, yet there is a uniting concept that behavior and its associations with important covariates are expressed as a function of near-continuous time.

The first talk applies TVEM to longitudinal panel data with developmental age as the time metric. This paper focuses on the prevalence of frequent heavy alcohol, cigarette, and marijuana use across nearly two decades spanning adolescence through young adulthood using a large, national sample. Results document moderating effects of gender and race/ethnicity as a function of continuous age. The second talk, which relies on intensive weekly diary data collected for up to 2.5 years, examines how contraceptive use changes over the course of young women’s sexual relationships. Factors associated with the desistence of condom use are examined as a function of time in relationship. The third talk uses repeated cross-sectional data from a national sample to estimate changes across historical time in the strength of the association between smoking and nicotine dependence. All three presenters will briefly show how the models were specified in SAS.

Together, these talks demonstrate the potential of TVEM to address novel questions about the role of time (including historical, developmental, and time from an event) in studying the epidemiology of health risk behaviors. This straightforward extension of multivariate regression analysis can aid in targeting and adapting intervention efforts to specific windows of time, and can reveal how policy changes made over historical time may have impacted rates and co-occurrence of behaviors. The discussant of this symposium is a distinguished researcher in the etiology and prevention of tobacco use and other health risk behaviors with extensive experience in innovative methods including TVEM. The discussion will focus on the potential TVEM has to be used in multiple areas of prevention research using diverse types of data in order to advance understanding of dynamic processes related to health risk behaviors.