Test-negative designs are commonplace in assessments of influenza vaccination effectiveness, estimating this value from the exposure odds ratio of vaccination among individuals treated for acute respiratory illness who test positive for influenza virus infection. This approach is widely believed to recover the vaccine direct effect by correcting for differential health-care-seeking behavior among vaccinated and unvaccinated persons. However, the relationship of the measured odds ratio to true vaccine effectiveness is poorly understood. We derived the odds ratio under circumstances of real-world test-negative studies. The odds ratio recovers the vaccine direct effect when 2 conditions are met: 1) Individuals' vaccination decisions are uncorrelated with exposure or susceptibility to the test-positive or test-negative conditions, and 2) vaccination confers "all-or-nothing" protection (whereby certain individuals have no protection while others are perfectly protected). Biased effect-size estimates arise if either condition is unmet. Such bias might suggest misleading associations of vaccine effectiveness with time since vaccination or the force of infection of influenza. The test-negative design could also fail to correct for differential health-care-seeking behavior among vaccinated and unvaccinated persons without stringent criteria for enrollment and testing. Our findings demonstrate a need to reassess how data from test-negative studies can inform policy decisions.