A0588
Title: Modeling developmental trajectories with nonrandomly missing data
Authors: Depeng Jiang - University of Manitoba (Canada) [presenting]
Abstract: Group-based trajectory analysis (GBTA) is commonly used for identifying distinctive developmental trajectories in longitudinal studies but may yield biased results when missing data is non-random. The aim is to compare conventional and extended GBTA in different missing data scenarios to understand their impact on trajectory identification. Simulation studies demonstrated that the extended GBTA outperformed the conventional GBTA in recovering trajectory estimates when classes were not well separated. In contrast, both methods were equally effective when classes were distinct. Applying these models to the Manitoba follow-up study data revealed four frailty trajectories, with age influencing trajectory membership. Marital status and living arrangements showed no significant associations. The extended GBTA's superiority is highlighted when handling non-random missing data, particularly in scenarios with unclear class distinctions. Valuable insights are provided for longitudinal studies with missing data, aiding researchers in understanding developmental heterogeneity and guiding future prevention strategies.
