Intensive care units (ICUs) are not always able to admit all patients who would benefit from intensive care. Pressure on ICU beds is likely to be particularly high during times of epidemics such as might arise in the case of swine influenza. In making choices as to which patients to admit, the key US guidelines state that significant priority should be given to the interests of patients who are already in the ICU over the interests of patients who would benefit from intensive care but who have not been admitted. We examine four reasons that in principle might justify such a prioritization rule and conclude that none is convincing. We argue that the current location of patients should not, in principle, affect their priority for intensive care. We show, however, that under some but not all circumstances, maximizing lives saved by intensive care might require continuing to treat in the ICU a patient already admitted rather than transferring that patient out of the unit in order to admit a sicker patient who would also benefit more from intensive care. We conclude that further modelling is required in order to clarify what practical policies would maximize lives saved by intensive care.