A Multilevel, Multimethod Longitudinal Study of Adolescent Health
Published in: American Journal of Preventive Medicine, v. 27, no. 2, Aug. 2004, p. 164-172
Posted on RAND.org on January 01, 2004
PURPOSE: To provide an overview of a multisite, long-term study that focuses on risk and protective factors, health behaviors (e.g., dietary practices, physical inactivity, tobacco use, and violent activity), and health outcomes (e.g., diabetes, obesity, and sexually transmitted diseases) for a fifth-grade cohort to be followed biennially from ages 10 to 20 years. METHODS: A two-stage probability sampling procedure was used to select 5250 fifth-grade students from schools in Birmingham AL, Houston TX, and Los Angeles CA to ensure a sufficient sample size of African Americans, Hispanics, and non-Hispanic whites, to support precise statistical inferences. Computer-assisted technology was used to collect data from children and their primary caregivers. Teachers and other school personnel responded to questionnaires, and observational procedures were used to obtain information about schools and neighborhoods. RESULTS: To exploit the multilevel, multimethod structure of the data, statistical models include latent-growth mixture modeling, multilevel modeling, time-series analysis, survival analysis, latent transition analysis, and structural equation modeling. Analyses focus both on the co-occurrence and predictors of growth trajectories for different health behaviors across time. CONCLUSIONS: By using a prospective research design and studying the predictors and time course of multiple health behaviors with a multilevel, multimethod assessment protocol, this research project could provide an empirical basis for effective social and educational policies and intervention programs that foster positive health and well-being during both adolescence and adulthood.