Assessing the Feasibility of Implementing Data Mining for Post-Licensure Vaccine Safety Surveillance Within the Vaccine Safety Datalink

• Activity Number: 176
• Type: Contributed
• Date/Time: Monday, July 30, 2012 : 10:30 AM to 12:20 PM
• Sponsor: Section on Statistical Learning and Data Mining
• Author(s): Susan Shortreed*+ and Jennifer Nelson and Onchee Yu and Do Peterson
• Companies: Group Health Research Institute and Group Health Research Institute and Group Health Research Institute and Group Health Research Institute
• Keywords: Vaccine Safety ; Post-licensure Surveillance ; Data Mining

Abstract

Data mining techniques are designed to extract previously unknown patterns from large data sources, and they are increasingly being used in medical contexts to detect unexpected associations, signals, or trends. Currently in the US, the Vaccine Adverse Event Reporting System (VAERS) is used to track spontaneously-reported adverse events by patients and health care providers, and data mining methods have been applied in VAERS to detect adverse events possibly associated with a vaccine. Yet there are considerable limitations inherent in both the quality and the depth of information that VAERS collects. The Vaccine Safety Datalink (VSD) is a population-based collaborative effort between the CDC and 10 managed health care organizations with access to detailed information on vaccination records and health outcomes on over nine million members. In this talk, we will describe a project within the VSD which aims to assess the feasibility of implementing data mining methods for monitoring vaccines to identify previously unknown adverse events. Specifically, we will introduce and discuss some practical issues that arise when applying data mining methods to large heath care data bases.