We appreciate the comments of Jackson and colleagues regarding possible residual confounding by health status as an explanation for the protection provided by influenza vaccination against death in our study.1 Confounding and biases represent ever-present threats to validity for even the best thought-out observational research.2 We controlled for numerous potential confounders, including health status, nursing home residence, and severity of illness on admission in multivariable models, and the observed protective effect of influenza vaccination remained robust. Furthermore, we included only hospitalized individuals with pneumonia. The implicit exclusion of vaccinated individuals at low risk of pneumonia-related hospitalization makes a “healthy vaccine effect” unlikely.
Fisman DN. To Rule Out Confounding, Observational Studies of Influenza Vaccine Need to Include Analyses During the “Preinfluenza Period”—Reply. Arch Intern Med. 2007;167(14):1554-1555. doi:10.1001/archinte.167.14.1554