Using Spatial Interaction Models to Track and Model the Spatial and Temporal Distribution of COVID-19: Case Studies from Central and East Europe (CEE) in the Third Year of the Pandemic

Abstract

An innovative approach was developed to forecast the spatiotemporal dynamics of COVID-19. This involved integrating global compartmental epidemic models with spatial interaction models using a metapopulation framework. The methodology was assessed using specific cases from distinct time periods in Poland, Germany, and the Czech Republic. The spatial interaction models incorporated mobility as a key determinant, estimated based on anticipated contact numbers, regional population, and resident distances. The spatial diffusion model enabled the temporal prediction of COVID-19 cases and the simulation of their spatial distribution, reflecting expected regional pandemic dynamics. This model facilitated predicting disease incidence in various counties and districts, following the NUTS geocoding standard. In 2022, the investigation spanned the entirety of Poland, Germany, and the Czech Republic at different times during the third year of the pandemic. Validation was performed using official epidemiological data from national sanitary epidemiology services. Analysis results revealed that integrating epidemiological models with spatial interaction models, particularly unconstrained gravity models and destination-constrained models, achieved a determination coefficient of nearly 90%. This high coefficient indicated a strong fit of the model to the spatiotemporal distribution of the validation data.