The test-negative design is often used to estimate vaccine effectiveness in influenza studies, and has recently been proposed in the context of Ebola outbreaks. It was introduced as a variation of the case-control design, in an attempt to reduce confounding bias due to healthcare-seeking behaviour. However, examining the directed acyclic graphs that describe the test-negative design reveals that, without strong assumptions, the estimated odds ratio under this sampling mechanism is not collapsible over the selection variable, such that the results obtained for the sampled individuals cannot be generalised to the whole population. We show in extensive simulations that logistic regression consistently overestimates vaccine effectiveness, and results in inflated type I error rates. To mitigate this, we propose a variation of the test-negative design -- the severity-adjusted test-negative design. Conditioning on severity of disease alleviates bias considerably, and under certain assumptions even makes it possible to unbiasedly estimate a causal odds ratio.