Optimizing Response Strategies to the COVID-19 Pandemic in Nova Scotia

Abstract

Epidemics require dynamic response strategies that encompass a multitude of policy alternatives to effectively balance health, economic, and societal considerations. Due to the complexity and wide array of policy alternatives, decision makers require tools to determine effective strategies and assess their impact. This thesis proposes a simulation-optimization framework to aid policymakers select closure, protection, and travel policies to minimize the total number of infections under a limited budget. The proposed framework combines a modified, age-stratified susceptible-exposed-infected-recovered (SEIR) compartmental model to evaluate the health impact of response strategies and two meta-heuristic optimization procedures, namely Genetic Algorithm (GA) and Simulated Annealing (SA), to effectively search for better strategies. Two types of response strategies are considered: time-based and state-based. The former is proactive in nature, prescribing at the outset response policies to be implemented over a set period of time, whereas the latter is a reactive strategy that adjusts the policies based on the number of new infections observed, mimicking the method often used by policymakers. Both frameworks were implemented on a real case study in Nova Scotia to devise optimized response strategies to COVID-19. The two approaches found a clear trade-off between health and economic considerations. The time-based results show regardless of the budget, policy makers should oscillate between policies of varying degrees of strictness. Closure policies seem to be the most sensitive to policy restrictions, followed by travel policies. Results suggest that after a budget threshold is met, practicing social distancing and wearing masks are always recommended. The state-based results set the optimal limits such that restrictions are tightened whenever there are signs of a potential community spread and loosened when the spread is contained. Given the high infectivity of the disease, the lower limits (triggering the shift to less strict policies) are set quite low, ranging between 0 and 20 infections depending on the budget and the existing policies. Both frameworks are generic and can be extended to encompass vaccination policies and to use different epidemiological models or optimization methods. The model was also used to compare potential policy scenarios during the vaccination period, to help determine a suitable timeline for lifting the restrictions while minimizing the public health impact. The test results indicate that masks and social distancing will be required in order to continue to keep the case count and hospitalizations low, even as closure and travel policies are relaxed.