Session: Time-Varying Effect Models: A Novel Technique for Addressing Crucial Questions in the Prevention of Health-Risk Behavior (Society for Prevention Research 23rd Annual Meeting)

Effective prevention of health-risk behaviors likely depends on a strong understanding of their timing. When are these behaviors most prevalent across historical, developmental, and day-to-day time? When do risk factors have their strongest effects? How long does it take for intervention effects to emerge, and how long do they last? How do these time-varying processes differ from person to person?

This poster forum shows how time-varying effect modeling (TVEM) can be used to answer these and other questions relevant to the prevention of health-risk behavior. The six posters presented are diverse in their application, but united in using TVEM to examine the ebb and flow of health-risk behavior processes across distinct yet interrelated time metrics.

Poster 1 uses TVEM in two nationally representative samples to show that substance use rates are declining over historical time, with some exceptions; and the peak risk for nicotine dependence coincides with first regular smoking at age 11.

Posters 2 and 3 show how risk factor effects change over developmental time. Poster 2 examines the dynamic relationship between violence exposure and substance use, showing that the relationship emerges later and lasts longer for females versus males. Poster 3 finds that initial nicotine dependence, which can occur soon after smoking initiation, predicts smoking frequency across adolescence, with its strongest effects appearing early.

Posters 4 and 5 use TVEM in two studies that intensively measured people’s intentions, moods, and behaviors across weeks and days, respectively. Poster 4 examines how sexual behaviors change since the start of a relationship, showing that initial intentions predicted sexual behavior during the entire first year, but only predicted condom use during the first 7 weeks. Poster 5 examines time-varying effects of smoking cessation treatment on anhedonia, a key predictor of relapse during cessation attempts. Treatment effects were complex functions of time, operating on both (a) mean levels of anhedonia and (b) the associations of anhedonia with negative affect and craving during the first two weeks after quitting.

Poster 6 introduces mixture TVEM (MixTVEM), an approach combining latent class growth modeling and TVEM to identify hidden subgroups that differ in their dynamic covariate functions across time. Using MixTVEM, the authors identified three classes of individuals who differed in the timing and magnitude of momentary relationships between negative affect and smoking urge following a quit attempt.

The forum will close with discussant Lisa Dierker, an expert in the application of novel statistical methods in health-risk behavior research. Discussion will focus on the implications of these findings for research, policy, and prevention.