Mechanism of Optimal Time-Course COVID-19 Vaccine Prioritization Based on Non-Markovian Steady-State Prediction

Vaccination is essential for controlling the coronavirus disease (COVID-19) pandemic. An effective time-course strategy for the allocation of COVID-19 vaccines is crucial given that the global vaccine supply will still be limited in some countries/regions in the near future and that mutant strains have emerged and will continue to spread worldwide. Both asymptomatic and symptomatic transmission have played major roles in the COVID-19 pandemic, which can only be properly described as a typical non-Markovian process. However, the prioritization of vaccines in the non-Markovian framework still lacks sufficient research, and the underlying mechanism of the time-course vaccine allocation optimization has not yet been uncovered. In this paper, based on an age-stratified compartmental model calibrated through clinical and epidemiological data, we propose optimal vaccination strategies (OVS) through steady-state prediction in the non-Markovian framework. This OVS outperforms other empirical vaccine prioritization approaches in minimizing cumulative infections, cumulative deaths, or years of life lost caused by the pandemic. We found that there exists a fast decline in the prevention efficiency of vaccination if vaccines are solely administered to a selected age group, which indicates that the widely adopted strategy to continuously vaccinate high-risk group is not optimal. Through mathematical analysis of the model, we reveal that dynamic vaccine allocations to combinations of different age groups is necessary to achieve optimal vaccine prioritization. Our work not only provides meaningful references for vaccination in countries currently lacking vaccines and for vaccine allocation strategies to prevent mutant strains in the future, but also reveals the mechanism of dynamic vaccine allocation optimization, forming a theoretical and modelling framework empirically applicable to the optimal time-course prioritization. Funding: This work was supported by the Hong Kong Baptist University (HKBU) Strategic Development Fund. This research was conducted using the resources of the High- Performance Computing Cluster Centre at HKBU, which receives funding from the Hong Kong Research Grant Council and the HKBU. Declaration of Interests: The authors declare that they have no competing interests.