Allocating COVID-19 vaccines : save one for the second dose?

The majority of COVID-19 vaccines currently available are typically administered in two doses, with a prescribed number of weeks separating the two doses. Because of uncertainty in vaccine supply, many vaccination managers have saved doses in inventory to ensure on-time second doses. However, saving doses slows the administration of first doses and potentially delays completing the vaccination of the target population. In this paper, we use a mathematical model to explore the performance of policies to manage the administration of two-dose vaccines given supply uncertainty. The mathematical structure of the model suggests simple ``set-aside'' policies that prioritize second doses and set aside some vaccine for second doses that are due in the coming weeks; any remaining vaccine is given as first doses. We can determine good set-aside policies using a simple simulation model. In our numerical experiments, the performance of the recommended set-aside policy is close to a bound generated by the optimal allocation given perfect information about vaccine supply; thus the set-aside policies are nearly optimal given uncertainty in supply. We also find that a "lockbox'' policy that ensures no second doses will be delayed appears to be overly conservative. We also show that set-aside policies perform well when vaccination supply is increasing and when two two-dose vaccines are available