Latent growth modeling of a nutrition and physical activity intervention
The health, economic, and social impacts of physical inactivity and unhealthy dietary patterns are quite significant and evidenced by the fact that only 3-4% of American adults follow all the dietary advice recommended by the Dietary Guidelines for Americans (DGA; Kohatsu, Robinson, & Torner, 2004), and specific subpopulations, including the rural South, are affected even more drastically (McCabe-Sellers et al., 2007). Furthermore, the majority of the people in the United States do not currently meet recommended amounts of physical activity (PA) and have not since the mid 1980s (U.S. Department of Health and Human Services [USDHHS], 2008). To address the discrepancy between health recommendations and actualized unhealthy patterns of physical activity and diet, community members in a small, urban area composed of a racially diverse population in Mississippi collaborated with university researchers to develop a community partnership that was named H.U.B. City Steps. The primary purpose of the present study was to use latent growth modeling (LGM) to determine relationships between latent variables and examine factors that not only prompt healthy behavior changes but allow for individual behavior maintenance within the H.U.B. City Steps sample. A secondary analysis used a structural equation model (SEM) to test interrelationships of variables that have been theorized to impact blood pressure. The LGM was unsuccessfully executed while the SEM revealed that none of the paths leading to systolic blood pressure (SBP) were significant. In the SEM, while the paths to SBP were not significant, significant relationships demonstrated between SES and the psychosocial latent variable as well as SES and physiologic variables in this primarily African American population are of interest. A model encompassing an individual's past health in conjunction with factors reflecting the cultural, social, and community context in which the individual is seated could be used to inform future interventions. Future research could examine the pathways of individual characteristics that form behaviors and then how the behaviors formed lead to health outcomes. The current LGM assessed individual factors and sought to predict relationships directly to health outcomes, but perhaps a more effective analysis might be to use health behaviors as the dependent variable and test health outcomes independently. While the primary analysis was unsuccessful, the research described herein is quite valuable. Future researchers need guidance as to how to approach conceptualizing and testing latent variables that are related to health behaviors. Furthermore, successful approaches, as well as unsuccessful modeling attempts, should be openly discussed for the benefit of future analysis.