Just ran across an interesting study. In it, Ruby Siddiqui argues that it's cost effective to treat all symptomatic patients with antiviral medications, even given the fact that you'll inevitably waste some courses on patients who aren't infected. What isn't cost-effective is to try and test each case and then administer antivirals. it's one of those cases where it's less expensive to permit some waste of the resource than it is to tightly identify the affected population. Abstract here:

Many countries are stockpiling antiviral (AV) drugs in preparation for a possible influenza pandemic. This study investigated the cost-effectiveness of such a strategy and the role of near-patient testing in conserving AV stocks, using a decision-analytical model. Under base-case assumptions (which included a fixed stockpile size that was smaller than the clinical attack rate) treating all symptomatic patients ('Treat Only') would be considered cost-effective (approximately £1,900-£13,800 per QALY gained, depending on the mortality scenario) compared with 'Do Nothing'. The 'Test-Treat' option (testing all symptomatic patients but treating only those that tested positive) resulted in moderate gains in QALYs at relatively large additional costs. The cost-effectiveness of 'Treat Only' was very sensitive to AV efficacy in preventing deaths. The model was also moderately sensitive to the time delay to the next pandemic and the discount rate. Other parameters (such as efficacy against complications, or the overall clinical attack rate) were not important determinants of the cost-effectiveness of this strategy...Stockpiling sufficient antivirals to treat all clinical cases appears to be a cost-effective strategy, provided antivirals are effective at preventing deaths from pandemic influenza. Stockpiling near-patient tests is not cost-effective.

A very similar argument, incidentally, can be made for stimulus spending.