Four Definitions of “Time” in Time-Varying Effect Modeling: Examples in Marijuana Use

Introduction. Time-varying effect modeling (TVEM), a statistical approach that enables researchers to estimate regression coefficients as complex functions of continuous time, holds enormous potential to advance prevention research. TVEM can address innovative questions about processes that unfold as a function of historical time, developmental age, age of onset, and time relative to a life event. We present a conceptual introduction to the approach and demonstrate four innovative ways to apply TVEM through empirical studies on the etiology of marijuana use.

Method. Study I examines changes in associations across historical time to understand how the link between marijuana use attitudes and marijuana use behavior has shifted from 1976 to present. Study II examines age-varying associations between heavy episodic drinking and marijuana use across developmental age in order to shed light on alcohol use as a risk factor with a dynamic effect. Study III explores the complex association between age of onset of marijuana use and adult marijuana use to identify precise age ranges during which the onset of use is most risky, and examines sex differences in this complex association. Study IV examines changes in marijuana use as a function of time relative to the birth of first child and how this trend differs across sex.

Results. Study I shows that marijuana use and attitudes follow a very similar nonlinear trend over historical time. The association between use and positive attitudes was strongest around the 1970s when rates were highest, but is weakening in recent years suggesting that attitudes may be a less important prevention target. Study II shows that rates of marijuana use peak around age 20 in this national sample, yet the strength of the association between heavy episodic drinking and marijuana use peaked during early adolescence. Study III demonstrates that early onset of marijuana use is a stronger risk for females compared to males. The risk for adult marijuana use is magnified for ages of onset through around age 18, suggesting that onset during mid- to late-adolescence should also be considered high risk. Study IV shows a developmental trend in marijuana use characterized by a peak in use at three years prior to birth of first child, followed by a steady decline through two years after the child’s birth. Females’ overall rates were lower, and their decrease during the four year period was more marked.

Conclusions. By conceiving of “time” in TVEM in different ways, critical new information can be gleaned from existing prevention and etiology data sets. TVEM holds great potential to advance prevention research by deepening our understanding of complex, dynamic processes.