This paper shows that if the relative risks of hospitalization of exposed compared to non-exposed to a putative study factor are different in comparison groups of in-patients - cases and non-cases of a specific disease - the odds-ratio estimate will be biased. This bias can be in any direction, either overestimated or underestimated, due to the selection factor of hospitalization. Finally, it is derived an algebraic expression that can correct the odds-ratio estimate.
