Objective To develop a rational dynamic drug allocation strategy in the early stage of emerging respiratory infectious disease pandemic when resource supplies are limited but more supplies are needed, minimize the fatal case count in the pandemic and provide scientific evidence for the resource allocation in the context of limited supply.

Method An infectious disease dynamics model of susceptible persons, latent infection case, untreated infection cases, treated infection cases, recovered persons death cases (SEIuItRD) was constructed based on age groups to analyze the impact of pharmacotherapy on recovery and death rates. The dynamic growth process of pharmacy production from shortage to saturation was simulated by using Logistic S-shaped function. The simulated 150-d pandemic period was divided into 15 decision-making stages, and a multi-stage nonlinear optimization model was constructed. The sequential quadratic programming algorithm was used to calculate the optimal allocation ratio for the comparison with the benchmark allocation strategy based on population proportion and the specific strategy for population at high death risk. Sensitivity analysis was conducted on key parameters.

Results Compared with the benchmark allocation strategy based on population proportion, the dynamic optimization strategy reduced the cumulative fatal cases by 16.10%. The optimal allocation plan exhibited a distinct "three-stage" characteristics: in the early stage of the pandemic, when the number of infections was small and the constraint of drug resources was relatively low, the drug allocation was made based on population proportion. In the incidence peak period of the pandemic, the drug resources were highly allocated to elderly group, while the allocation proportions for children and adults were reduced to zero, then 12.36% of the allocation was made to adult group and 87.64% to elderly group from day 90 to day100 , 3.00% of the allocation was made to children group, 94.39% to adult group, and 2.61% to elderly group from day 100 to day 110. In the late stage of the pandemic, the allocation ratio was adjusted to that based on population proportion. The sensitivity analysis results showed that as the drug efficacy increased from 0.50% to 0.95%, the cumulative fatal cases under all the three strategies decreased. When the pharmacy production growth rate (k) increased from 0.01 to 0.10, the cumulative fatal cases under all the three strategies showed significant decreases. When k reached 0.07, the specific strategy for population at high death risk outperformed the benchmark strategy and the dynamic optimization strategy.

Conclusion In the early stage of a respiratory infectious disease pandemic with limited drug supply, a dynamic drug allocation strategy based on multi-stage optimization could significantly reduce the fatal cases caused by the pandemic. Adjusting the allocation plan dynamically according to changes in pharmacy production capacity is an effective way to maximize public health benefits.