Using the Time-Varying Effect Model with Intensive Longitudinal Data to Inform Prevention Research

Introduction: Designing effective intervention programs can involve an understanding of several aspects of the timing of processes, such as what periods are optimal for intervention and how long program effects last. This paper will discuss how the time-varying effect model (TVEM) can be applied to intensive longitudinal data (ILD) to uncover new information for creating targeted or tailored interventions. TVEM is a flexible, semiparametric model that examines how the associations between variables change in strength over continuous time.

This paper will examine how TVEM has a number of potential applications for prevention research. For example, it has been applied to understand processes unfolding in a smoking cessation study after a quit attempt using EMA data (Lanza, Vasilenko, Liu, Li, & Piper, 2014; Vasilenko, Piper, Lanza, Liu, Yang, & Li, 2014). Such an approach can identify important periods in which additional booster sessions or treatments may be added to increase program effects. In addition, TVEM can be used to identify which age or period to target for an intervention, which mediating factors to intervene on, and what messages may be beneficial for different subgroups at different periods of time (Coyle & DiClemente, 2014; Vasilenko & Lanza, 2014).

Methods and Results: As an empirical example, I present TVEM analyses of intensively collected weekly data from young women over 2.5 years (N=587; race = 59.5% white, 33.9% Black or African American, 5.3% Native American or Alaskan native, 1.4% Asian/Native Hawaiian or other Pacific Islander; ethnicity = 8.5% Hispanic; Mage = 18.6, SD = .62). Analyses examine rates of sexual behavior and condom use over time in a relationship with a sexual partner, and how baseline characteristics (e.g., demographic factors, personal attitudes) differentially predict these behaviors over time. For example, compared to white women, young black women were more likely to engage in sexual behavior during the 30^th-40^th week in a relationship with a partner, but those who were sexually active were also more likely to use condoms during the first ten weeks of a new relationship. Results could be applied to partner-focused intervention programs to prevent risky sexual behavior, and could help in the development to interventions tailored to different subgroups at different stages of relationships.