Generating Policy-Relevant Statistical Evidence in Sequentially Monitored Vaccine and Drug Safety Evaluations Using Electronic Health Record Data

• Activity Number: 465
• Type: Invited
• Date/Time: Wednesday, August 3, 2016 : 8:30 AM to 10:20 AM
• Sponsor: Health Policy Statistics Section
• Author(s): Jennifer Clark Nelson* and Andrea J. Cook and Robert Wellman and Ram Tiwari and Michael Nguyen and Estelle Russek-Cohen and Tracey Marsh and Azadeh Shoaibi and Denise Boudreau
• Companies: Group Health Research Institute and Group Health Research Institute and Group Health Research Institute and FDA/CDER/OT/OB and FDA and FDA and University of Washington and FDA and Group Health Research Institute
• Keywords: : electronic health record (EHR) data ; health policy statistics ; safety surveillance ; sequential testing

Abstract

Proactive drug and vaccine safety surveillance has been made possible by the creation of large health care data networks combined with routine sequential monitoring of the electronic data they contain. To conduct sequential testing in this observational environment, existing trial-based approaches have been extended to incorporate confounders, accommodate rare events, and address data privacy constraints that prevent individual-level data pooling. Most adaptations for this new setting have involved design-based confounder strategies (e.g., matching, stratification), while analysis-based approaches (e.g., regression, weighting) have received less attention. Methods have also typically focused on relative comparisons of risk, despite heavy reliance by policy-makers on risk differences for decision-making. We describe two newer group sequential approaches for health care data networks that implement analysis-based confounder adjustment, including a risk difference estimation approach. We illustrate the methods using data from FDA's Sentinel network and comment on challenges and opportunities for future statistical contributions in this emerging field of post-market regulatory science.