American Journal of Preventive Medicine
April 2015 Volume 48, Issue 4, p365-490
http://www.ajpmonline.org/current

Medicare Claims Versus Beneficiary Self-Report for Influenza Vaccination Surveillance
Kimberly A. Lochner, ScD, Marc A. Wynne, MSPH, Gloria H. Wheatcroft, MPH, Chris M. Worrall, BS, Jeffrey A. Kelman, MD
Published Online: February 17, 2015
DOI: http://dx.doi.org/10.1016/j.amepre.2014.10.016
Abstract
Background
Although self-reported influenza vaccination status is routinely used in surveillance to estimate influenza vaccine coverage, Medicare data are becoming a promising resource for influenza surveillance to inform vaccination program management and planning.
Purpose
To evaluate the concordance between self-reported influenza vaccination and influenza vaccination claims among Medicare beneficiaries.
Methods
This study compared influenza vaccination based upon Medicare claims and self-report among a sample of Medicare beneficiaries (N=9,378) from the 2011 Medicare Current Beneficiary Survey, which was the most recent year of data at the time of analysis (summer 2013). Sensitivity, specificity, positive predictive value, and negative predictive value were calculated using self-reported data as the referent standard. Logistic regression was used to compute the marginal mean proportions for whether a Medicare influenza vaccination claim was present among beneficiaries who reported receiving the vaccination.
Results
Influenza vaccination was higher for self-report (69.4%) than Medicare claims (48.3%). For Medicare claims, sensitivity=67.5%, specificity=96.3%, positive predictive value=97.6%, and negative predictive value=56.7%. Among beneficiaries reporting receiving an influenza vaccination, the percentage of beneficiaries with a vaccination claim was lower for beneficiaries who were aged <65 years, male, non-Hispanic black or Hispanic, and had less than a college education.
Conclusions
The classification of influenza vaccination status for Medicare beneficiaries can differ based upon survey and claims. To improve Medicare claims–based surveillance studies, further research is needed to determine the sources of discordance in self-reported and Medicare claims data, specifically for sensitivity and negative predictive value.