Active Influenza Vaccine Safety Surveillance: Potential Within a Healthcare Claims Environment.
Brown, Jeffrey S. PhD *+; Moore, Kristen M. MPH *+; Braun, M Miles MD, MPH ++; Ziyadeh, Najat MA, MPH [S]; Chan, K Arnold MD, ScD [S][P]; Lee, Grace M. MD, MPH *[//]; Kulldorff, Martin PhD *; Walker, Alexander M. MD, DrPH [P]**; Platt, Richard MD, MSc *+
47(12):1251-1257, December 2009.
(Format: HTML, PDF)
Background: Rapid safety assessment of novel vaccines, especially those targeted against pandemic influenza, is a public health priority.
Objectives: Assess the feasibility of using healthcare claims data to rapidly detect influenza vaccine adverse events using sequential analytic methods.
Research Design: Retrospective pilot study simulating prospective surveillance using 6 cumulative monthly administrative claims data extracts. The first included encounters occurring in October; each subsequent extract included an additional month of encounters. Ten adverse events were evaluated, comparing postvaccination rates during the 2006-2007 influenza season to those expected based on rates observed in the prior season.
Subjects: Members of a large, multistate health insurer who had a claim for influenza vaccination during the 2005-2006 or 2006-2007 influenza seasons.
Measures: The completeness of monthly claims extracts.
Results: Most vaccinations and outcomes were identified early in the 2006-2007 season; about 50% of vaccinations and short latency events were identified in the second monthly data extract, which would typically become available by mid-December, and 80% of vaccinations and events were identified in the third extract. With respect to overall claims lag, approximately 90% of vaccinations and events were identified within 1 to 2 months after vaccination, regardless of vaccination month.
Conclusions: This study suggests that administrative claims data might contribute to same season influenza vaccine safety surveillance in large, defined populations, especially during a threat of pandemic influenza. Based on our previous work, we believe this method could be applied to multiple health plans' data to monitor a large portion of the US population.
(C) 2009 Lippincott Williams & Wilkins, Inc.