Piecewise growth curve modeling approach for longitudinal prevention study

Conventional growth curve modeling typically involves a single growth profile to represent changes in an outcome variable across time. This study applies an alternative growth curve model with multiple profiles to incorporate multiple developmental stages from which longitudinal data have been obtained. Identifying and modeling the multiple transitional stages involved in repeated measures are both theoretically and empirically important in longitudinal study. A longitudinal data set obtained from a substance use prevention study was used as an example to illustrate this application. The data contained a total of 50 junior high schools (23 control and 27 program schools) observed for seven waves crossing junior high school and high school stages, with the prevalence rate of monthly cigarette use as the outcome measure. Comparisons showed that the piecewise growth curve models incorporating multiple stages demonstrated significant improvement on model fitting compared to the single-piece growth curve model. Results showed marginal prevention effects in the junior high school stage but not in the high school stage. Piecewise growth curve models offer more both substantively and analytically appropriate model specification and flexibility in incorporating transitional periods in studying changes across time. (C) 2003 Elsevier B.V. All rights reserved.