American Statistical Association
We present a novel repeated latent class model for a longitudinal categorical outcome that is frequently measured as in our prospective study of older adults with monthly data on activities of daily living (ADL) for more than ten years. The longitudinal response is easy to obtain and measured at much higher frequency than other covariates. We propose to use repeated latent classes to capture distinct trajectory patterns that may repeat themselves or change over time. Individuals may remain in a trajectory class or switch to another trajectory class over time as the class membership predictors are updated.
The identification of a common set of trajectory patterns across time allows periodic changes to be distinguished from local fluctuations in the response. An informative event of interest such as death can be jointly modeled by trajectory class-specific probability of the event through shared random effects. We do not impose the conditional independence assumption that is typically assumed in prior latent class modeling. The method is illustrated by analyzing the change over time in ADL trajectory class among 754 older adults with 70500 person-months of follow-up in the Precipitating Event Project. We also investigate the impact of modeling the class-specific probability of death on the estimates of ADL trajectory classes in a simulation study. Allowing periodic updating of trajectory classes for a longitudinal categorical response without assuming conditional independence highlights our contribution.
|Date:||Tuesday, October 23, 2012|
|Time:||3:30 - 4:30 P.M.|
New York State Psychiatric Institute
1051 Riverside Drive
Pardes Building, Multipurpose Room 6602
New York, New York