Influenza vaccine effectiveness among patients on hemodialysis: methods to control the healthy-user bias Public Deposited

Downloadable Content

Download PDF
Last Modified
  • March 21, 2019
  • McGrath, Leah Jill Sirkus
    • Affiliation: Gillings School of Global Public Health, Department of Epidemiology
  • Background: Patients with end-stage renal disease (ESRD) are at increased risk for several preventable infections. Although vaccines have been recommended for people with ESRD for many years, little is known about the level of vaccine effectiveness (VE) in preventing clinical health outcomes in this population. Observational studies of VE are challenging, however, because vaccinated persons may be healthier than unvaccinated persons. This dissertation aims to estimate influenza vaccine effectiveness among patients on hemodialysis using novel ways to control for bias. Methods: Using Medicare claims from patients on hemodialysis, a natural experiment was created by using year-to-year variation in the match of the influenza vaccine to the circulating virus. VE for influenza-like illness, influenza/pneumonia hospitalization, and mortality was estimated by comparing matched (1998, 1999, 2001) and mismatched (1997) years among vaccinated patients. An alternate method identified time-varying predictors of vaccination status and used these variables to control for time-varying confounding using a marginal structural model. Results: Conventional estimates comparing vaccinated patients with unvaccinated patients suggested a large protective effect--influenza vaccine reduced mortality by 30%. The pooled VE estimate from the natural experiment of comparing matched seasons to a placebo was 0% (95% CI: -3,2%) for influenza-like illness, 2% (95% CI: -2,5%) for hospitalization, and 0% (95% CI: -3,3%) for death. Hospitalization and skilled nursing care were highly associated with not being vaccinated, suggesting that these variables could be used to control for the healthy-user bias. The marginal structural model remained biased even after accounting for time-varying confounding, which likely resulted in exaggeration of the protective effect of the vaccine. Conclusions: Strong confounding bias is present when estimating influenza vaccine effectiveness. Controlling for bias using a natural experiment resulted in estimates of VE for all outcomes that were close to the null. This suggests that the current influenza vaccine strategy may have a smaller effect on morbidity and mortality in the ESRD population than previously thought. Alternate strategies (high dose vaccine, intradermal vaccine, and adjuvanted vaccines) should be investigated to achieve better health outcomes.
Date of publication
Resource type
Rights statement
  • In Copyright
  • Brookhart, M. Alan
  • Doctor of Philosophy
Degree granting institution
  • University of North Carolina at Chapel Hill
Graduation year
  • 2012

This work has no parents.