How viruses spread across space and time: forecasting pandemic progression by modelling geographico-temporal interactions

The COVID-19 pandemic has revealed severe flaws in the global healthcare systems ability to respond to unexpected health catastrophes. Much of the confusion and mishandling of the situation could be attributed to the failure in accurately predicting the spread of the virus across geographical locations. A global resource shortage in essential medical supplies and equipment, such as personal protective equipment (PPE) and ventilators, led to a compromised global supply chain. As a result, resources could not be allocated as needed to curb the spread of the pathogen in the most efficacious way. Although forecast models and machine learning algorithms have served as invaluable tools in devising effective response strategies, a large majority of these models were limited by their ability to describe the intricate interactions that underlie the spatio-temporal dynamics of viral proliferation. To address this issue, we employed a vector autoregression model to help capture the evolution of the disease across both the spatial and the temporal axes. Unlike traditional autoregression models, the present model is able to account for statistical regularities that exist both within a given region, and between geographical locations. Our results demonstrate that this approach accurately described the relationships across domestic and international localities throughout the evolution of the disease.