There is currently considerable uncertainty about the effect of the influenza vaccine on the risk of mortality and other outcomes among community-dwelling elderly individuals. Traditional epidemiologic cohort studies comparing vaccinated with unvaccinated elderly subjects have attributed very large protective effects to receipt of the vaccination.1 However, these studies appear to be strongly biased by “unmeasured confounding,” that is, unobserved or uncontrolled differences between the groups. Even after statistical control for various demographic variables and other medical conditions, elderly patients who receive the vaccine appear to be much healthier than apparently comparable unvaccinated patients. Strong evidence of this bias comes from an important epidemiologic study by Jackson et al2 in which the authors reported that a large beneficial effect of the vaccine is apparent before the influenza season begins.