A Bayesian Self-Controlled Case Series Method for Large-Scale Longitudinal Data in Drug Safety Surveillance

• Activity Number: 637
• Type: Contributed
• Date/Time: Thursday, August 5, 2010 : 8:30 AM to 10:20 AM
• Sponsor: Section on Statistical Learning and Data Mining
• Author(s): Shawn Evelyn Simpson*+ and David Madigan
• Companies: Columbia University and Columbia University
• Keywords: drug safety ; Bayesian statistics ; self-controlled case series

Abstract

Many post-approval drug safety surveillance methods are based on 2 x 2 tables of drug-condition pairs. Such analyses ignore drug interactions and have the potential to give misleading results. We extend the self-controlled case series (SCCS) method (Farrington, 1995) for longitudinal health records to provide a new procedure for large-scale post-marketing drug safety surveillance. This model implicitly controls for fixed baseline covariates since each individual acts as their own control. We expand the model to include large numbers of potentially time-varying confounders such as other drugs. Our Bayesian version of the SCCS method deals with high dimensionality and provides a sparse solution via a Laplacian prior. We present details of the model and optimization procedure, as well as empirical results for drugs and conditions that have been previously linked.