Modeling of Exposure Data from Point Sources Under a Matched Case-Control Design with Multiple Disease States

• Activity Number: 479
• Type: Contributed
• Date/Time: Wednesday, August 3, 2011 : 8:30 AM to 10:20 AM
• Sponsor: ENAR
• Author(s): Shi Li*+
• Companies: University of Michigan
• Keywords: matched case-control study ; point source modeling ; disease sub-classification ; Bayesian inference ; Markov chain Monte Carlo (MCMC) ; conditional likelihood

Abstract

In this paper, we extend the distance-odds models of Diggle et al (2000) proposed to investigate elevated risk/odds around point sources under a matched case-control design to distance-odds models where there are sub-types within cases in a matched design. We consider models analogous to the polychotomous logit models and adjacent-category logit models. Multiple point sources as well as covariate adjustments are considered. Maximum likelihood, profile likelihood, iteratively re-weighted least squares and a hierarchical Bayesian approach using Markov chain Monte Carlo were evaluated to conduct inference under these models. We compare these methods via an extensive simulation study and show that with multiple parameters and a nonlinear model, Bayesian methods have advantages in terms of estimation stability, precision and inference. The methods are illustrated by analyzing Medicaid claims data in pediatric asthma population in Detroit.