Conditional Trajectory Class Modeling for Bivariate Outcomes

• Activity Number: 251
• Type: Contributed
• Date/Time: Monday, August 5, 2013 : 2:00 PM to 3:50 PM
• Sponsor: Health Policy Statistics Section
• Author(s): Shu-Xia Li*+ and Haiqun Lin and Xiao Xu and Harlan Krumholz
• Companies: Yale University and Yale University and Yale University and Yale University
• Keywords: Bivariate outcomes ; condtional latent classes

Abstract

We develop a conditional trajectory class modeling approach for bivariate outcomes over time. We illustrate our method through analying dual hospital trajectories of risk-standardized cost and mortality rate. In the current group-based trajectory models of Jones and colleagues that are implemented in SAS proc TRAJ, either joint trajectory patterns of the two outcomes with one-to-one correspondence or two independent sets of trajectory classes (followed by cross-tabulated trajectory classes in both outcomes) are derived. Our proposed conditional approach, in contrast, identifies distinct trajectory classes for mortality rates, and then simultaneously identify distinct classes of cost trajectories using mortality trajectories as latent predictors. As a result, one or more mortality trajectories are allowed to share the same cost trajectory, and a same mortality trajectory is also allowed to have different cost trajectories. This allows us to account for the dependence between the two outcomes while without restricting them to a one-to-one correspondence. Association of hospital characteristics with the trajectory classes of each or the dual trajectory classes can be established.