Operating Characteristics of Statistical Methods for Active Drug Safety Surveillance

• Activity Number: 488
• Type: Invited
• Date/Time: Wednesday, August 3, 2011 : 10:30 AM to 12:20 PM
• Sponsor: IMS
• Author(s): Patrick Ryan*+
• Companies: Johnson & Johnson Pharmaceutical R&D, LLC

Abstract

The Observational Medical Outcomes Partnership (OMOP, http://omop.fnih.org) is a public-private partnership conducting methodological research to inform the appropriate use of stastical methods when applied to observational healthcare databases for active drug safety surveillance. OMOP designed an experiment to study the performance of different active surveillance methods against an array of disparate observational databases, in their ability to identify known drug associations and discern from false positive findings. Thirteen methods were implemented as alternative analysis approaches for active surveillance, including cohort and case-based designs, sequential testing approaches and multivariate models used in other data mining applications. These methods were executed across the OMOP data network of 10 data sources, both administrative claims and electronic health records, and run for a set of drug-outcome test cases with a defined ground truth. The resulting estimates were used to measure the operating characteristics (e.g. sensitivity, specificity) of the methods across the individual sources and within composite estimates obtained from meta-analysis.