The choice of therapy for febrile neutropenia is complex, because of the large number of options that are similar in safety and efficacy. However, there are a number of outcomes that may be useful when these choices have to be made. It is generally agreed that infection-related mortality is too rare an event, with the availability of modern antibiotics, to be of general use in treatment choices. Response to initial therapy may be useful, but differences among regimens in recent randomized trials only occasionally reach statistical significance, despite adequate power and sample size. The time to clinical response has been shown to vary significantly among otherwise similar regimens and may be very useful when response-based choices are made. Ideally, clinical and policy decisions should be based on a combined evaluation of outcomes and cost. In the case where clinical outcomes are the same for more than one regimen, cost-minimization analysis is appropriate. In the case where clinical outcomes differ, cost-effectiveness or cost-utility is an appropriate measure on which to base decisions. The cost of therapy can be easily estimated by using the number and average cost of days of hospitalization as a surrogate. A decision-analytic model for febrile neutropenia is proposed.