Abstract
During pandemics, such as COVID-19, supplies of vaccines can be insufficient to meet all needs, particularly when vaccines first become available. Our study develops a dynamic methodology for vaccine allo- cation, segmented by region, age, and timeframe, using a time-sensitive, age-structured compartmental model. Based on the objective of mini- mizing a weighted sum of deaths and cases, we used the Sequential Least Squares Quadratic Programming method to search for a locally opti- mal COVID-19 vaccine allocation for the United States, for the period from December 16, 2020, to June 30, 2021, where regions corresponded to the 50 states in the United States (US). We also compared our solu- tion to actual allocations of vaccines. From our model, we estimate that approximately 1.8 million cases and 9 thousand deaths could have been averted in the US with an improved allocation. When case reduction is prioritized over death reduction, we found that young people (17 and younger) should receive priority over old people due to their potential to expose others. However, if death reduction is prioritized over case reduc- tion, we found that more vaccines should be allocated to older people, due to their propensity for severe disease. While we have applied our methodology to COVID-19, our approach generalizes to other human- transmissible diseases, with potential application to future epidemics.