Analysis of relative effectiveness of high-dose versus standard-dose influenza vaccines using an instrumental variable method

Background: Observational studies of the relative effectiveness of influenza vaccines are essential for public health decision making. Their estimates, however, are subject to bias due to unmeasured confounders. Instrumental variable (IV) methods can control for observed and unobserved confounders. Methods: We used linked electronic medical record databases in the Veterans Health Administration (VHA) as well as Medicare administrative files to examine the relative vaccine effectiveness (rVE) of high-dose influenza vaccine (HD) versus standard-dose influenza vaccines (SD) in preventing hospitalizations among VHA-enrolled Veterans ≥65 years of age during 5 influenza seasons (2010–2011 through 2014–2015). Using multivariable IV Poisson regression modeling to address unmeasured confounding and bias, we analyzed the data by each season and through longitudinal analysis of all five seasons. Findings: We included 3,638,924 person–influenza seasons of observation where 158,636 (4%) were among HD vaccine recipients and 3,480,288 (96%) were among SD vaccine recipients. Of the 1,728,562 Veterans, 1,702,824 (98.5%) were male and 1,299,412 (75%) were non-Hispanic white. Based on the longitudinal analysis of all five seasons, the IV-adjusted rVE estimate of HD vs. SD was 10% (95% CI, 8–12%) against all-cause hospitalization; 18% (95% CI, 15–21%) against cardiorespiratory-associated hospitalization; and 14% (95% CI, 6–22%) against influenza/pneumonia-associated hospitalization. The findings by season were similar. Interpretation: Our analysis of VHA clinical data collected from approximately 1.7 million Veterans 65 years and older during five seasons demonstrates that high-dose influenza vaccine is more effective than standard-dose influenza vaccines in preventing influenza- or pneumonia-associated hospitalizations, cardiorespiratory hospitalizations, and all-cause hospitalizations.