Abstract:
|
In alcoholism research, two complementary outcomes are of interest: frequency of drinking (proportion of drinking days (PDD)) and intensity of drinking (average number of drinks per drinking day) during a specific timeframe. The two outcomes are often measured repeatedly over time on the same subject, and although they are closely related, they are rarely modeled together. We propose a joint model that fits these longitudinal outcomes simultaneously, using correlated random effects to model the association between the two outcomes and among repeated measurements within a subject. The model has two components: a hurdle binomial component for PDD, which is often zero-inflated, and a lognormal component for drinking intensity. Due to the computational impracticality of fitting models with a large number of random effects using standard frequentist approaches, we use a Markov Chain Monte Carlo-based Bayesian approach to fit the model. We illustrate the approach on data from COMBINE, a large clinical trial of alcoholism treatments. We conduct a simulation study to examine gains in parameter estimate bias and MSE associated with modeling the two outcomes jointly rather than separately.
|