Modeling Trajectories of Delirium Using Mixture Models

• Activity Number: 640
• Type: Contributed
• Date/Time: Thursday, August 5, 2010 : 8:30 AM to 10:20 AM
• Sponsor: Section on Risk Analysis
• Author(s): Long Huu Ngo*+ and Edward R. Marcantonio
• Companies: Beth Israel Deaconess Medical Center/Harvard Medical School and Beth Israel Deaconess Medical Center/Harvard Medical School
• Keywords: Delirium ; Trajectory ; Longitudinal ; Survival Analysis ; Mixture Models

Abstract

Delirium is associated with nosocomial infections and in-hospital mortality. We explored a longitudinal measure called MDAS (Memorial Delirium Assessment Scale) as a potential predictor of survival time. Each of the 455 patients in our study was evaluated using MDAS at 6 time points. Using a mixture model composed of two parts, the multinomial distribution of the class of trajectory, and the temporal shape of the trajectory modeled with a polynomial function, we obtained the likelihood function of the conditional distribution of the observed trajectory. We implemented six potential classes of trajectories, and 3 different temporal shapes (linear, quadratic, cubic). We selected the optimal mixture model using the Bayesian Information Criterion (BIC). We found the trajectory modeling approach using mixture models to be useful in relating the longitudinal MDAS data to survival time.