The beds of an intensive care unit (ICU) are a scarce resource. Stochastic
patient demands for these beds and stochastic service times in their utiliz
ation make managing that resource a complex problem lacking an easy solutio
n. The current practice in one Hong Kong hospital is for the ICU administra
tor to exploit the fact that there are some patients whose admission to the
unit can be postponed. These are patients scheduled for an elective surger
y that can be cancelled. One way to minimize the number of cancelled surger
ies is to reserve some of the unit's beds for the exclusive use of the elec
tive-surgery patients. We evaluate various bed-reservation schemes via a si
mulation model that is based on this ICU's historical data, and demonstrate
the tradeoffs that each requires among various relevant system-performance
measures. We further show how this information can be summarized in a clas
sic efficient frontier. This frontier provides a useful medium through whic
h the ICU administrator can communicate the rationale behind the chosen bed
-allocation system to the surgeons and the ICU physicians, in an attempt to
resolve the potential conflicts between them. (C) 2000 Elsevier Science B.
V. All rights reserved.