Spatial Allocation of Scarce Vaccine and Antivirals for COVID-19

Abstract Although the COVID-19 disease burden is heterogeneous across space, the U.S. National Academies of Sciences, Engineering, and Medicine recommends an equitable spatial allocation of pharmaceutical interventions based, for example, on population size, in the interest of speed and workability. Utilizing economic–epidemiological modeling, we benchmark the performance of ad hoc allocation rules of scarce vaccines and drugs by comparing them to the rules for a vaccine and for a drug treatment that minimize the economic damages and expenditures over time, including a penalty cost representing the social costs of deviating from an ad hoc allocation. Under different levels of vaccine and drug scarcity, we consider scenarios where length of immunity and compliance to travel restrictions vary, and consider the robustness of the rules when assumptions regarding these factors are incorrect. Because drugs and vaccines attack different points in the disease pathology, the benefits from deviating from the ad hoc rule differ. For drug treatment, optimal policies often allocate all available treatments to one jurisdiction for a period of time, while ad hoc rules act to spread out treatments across jurisdictions. For vaccines, the benefits from deviating are especially high when immunity is permanent, when there is compliance to travel restrictions, and when the supply of vaccine is low. Interestingly, a lack of compliance to travel restrictions pushes the optimal allocations of vaccine towards the ad hoc and improves the relative robustness of the ad hoc rules, as the mixing of the populations reduces the spatial heterogeneity in disease burden. JEL Classification C61, H12, H84, I18, Q54 Significance and Key Takeaways Contribution Much of the discussion around pharmaceutical interventions for COVID-19 has focused on what segments of the population to prioritize for vaccines. Two important questions have received much less attention but could be just as significant in reducing the economic and public health costs of COVID-19. First, there is the question of how much of a limited vaccine to allocate across jurisdictions within a country and within each jurisdiction. Should we allocate based on relative population size, some normalized measure of disease burden (e.g., cases per 100,000), or should we base the allocation on numbers of essential workers? Second, other pharmaceutical interventions, such as antiviral drugs are coming online and will be in limited supply at least initially. How should these drugs be allocated across and within States? We investigate these questions by benchmarking the economic and public health performance of ad hoc allocation rules against optimally-derived rules. We also investigate how robust allocation rules are to compliance to nonpharmaceutical interventions, such as travel restrictions, and to uncertainty on the degree of immunity conveyed by the pharmaceutical interventions and prior infection. Knowledge on compliance to travel restrictions is critically important to the allocation question, as research has shown that varying levels of compliance across jurisdictions impacts the spatial distribution of disease burden. Whether the economic and public health implications from spatial allocation rules are as significant as allocation within jurisdictions to different classes of people is an open question and likely depends on the objectives of the policymakers. We find economic and public health gains from spatially targeted rules even after considering additional costs associated with deviations from the ad hoc rules. Antiviral Drugs Spatial Prioritization: Drugs should be directed towards the region that has fewer infections; optimal allocation gives rise to extreme allocations where it is preferable to give all of the allotment to one jurisdiction for a period of time rather than an allocation based on relative levels of infected individuals. Epidemiological Consequences of Spatial Prioritization: Compared to the ad hoc allocation rule, where more treatments go to locations with more infected individuals, the optimal allocation results in a higher number of cumulative cases in the jurisdictions that have a higher initial level of infected individuals and fewer cumulative cases in the jurisdictions that have a lower initial level of infected individuals. Over the period of four months, however, how well the optimal more extreme allocation does in aggregate (across jurisdictions) relative to an aggregate ad hoc allocation is dependent on epidemiological, behavioral, and logistical factors. Robustness of Spatial Allocation: The optimal allocations are not robust to incomplete information on compliance to travel restrictions and immunity, and perform worse than the ad hoc allocations in cases where they are designed under one set of assumptions but yet the true state of the world is different. Policy Recommendation: Until we have more information about compliance and immunity, our analysis leads us to conclude that ad hoc allocations may be the least risky option for the allocation of antiviral drugs. Vaccines Spatial Prioritization: Compared to an ad hoc allocation rule based on relative population size as recommended by the US National Academies of Science, Engineering, and Medicine, the optimal allocation of vaccine favors the least-burdened jurisdiction, resulting in an unequal distribution from a resource allocation perspective. Epidemiological Consequences of Spatial Prioritization: The optimal allocation results in a more equal level of infection across jurisdictions in each period while the ad hoc allocation results in a more equal distribution of aggregate cumulative infections. Robustness of Spatial Allocation: In terms of economic expenditures and cumulative cases, the optimal allocation is less impacted when assumptions on immunity are incorrect than when wrong about compliance to travel restrictions. Policy Recommendation: Imposing strict travel restrictions, or e.g. forcing quarantine when traveling to another jurisdiction, and prioritizing vaccination in jurisdictions that have lower initial disease burden could prevent a significant number of cases.