Latent Transition Analysis to Classify the Food Environment and It’s Effect on Health Outcomes

• Activity Number: 289 - Contributed Poster Presentations: Section on Statistics in Epidemiology
• Type: Contributed
• Date/Time: Tuesday, August 9, 2022 : 10:30 AM to 12:20 PM
• Sponsor: Section on Statistics in Epidemiology
• Author(s): Kelsey Alexovitz* and Brisa Sanchez and Emma Sanchez-Vaznaugh
• Companies: Drexel University and Drexel University and San Francisco State University
• Keywords: Latent Transition Analysis; Zero-Inflated Poisson; Built Environment; Food Environment

Abstract

Examining how health outcomes are influenced by the food environment (FE) is challenging given that multiple features (e.g. different types of food outlets) can be used to measure the FE, and that the features vary in spatial proximity to subjects and over time. Moreover, measures of food environment features have an excess of zeros. We propose a latent transition analysis (LTA) model to classify the food environment that uses zero-inflated Poisson distributions to model the observed multivariate count data. Using simulations, we examine how our approach compares to classical LTA that uses multinomial distributions for observed variables. We apply our modeling strategy to classify the food environment near California public schools. Using this classification, we examine the association between time varying school neighborhood food environment types and longitudinal measures of pupils’ body weight. The derived classification of the food environment near schools can be used to inform interventions and other preventative strategies.