A probabilistic approach for situation awareness and forecasting of the COVID-19 pandemics in Norway

During the COVID-19 pandemic, governments have implemented interventions to control the spread of the SARS-CoV-2 virus. Often, decisions have been taken based on the assumptions of homogeneity of the pandemics, across local regions and time, also because of the lack of mathematical and computational instruments that can capture and predict local, temporal dynamics.

In this talk, Dr Arnoldo Frigessi and Dr Birgitte Freiesleben de Blasio will present a probabilistic model which estimates regional and local transmission in their complex interactions in space and time. Their model is a regional metapopulation model, informed by daily updated real-time mobile phone mobility data and multiple epidemical time series, including hospital admission and positive and negative test data.

Both propose a new inferential method, the split sequential Monte Carlo Approximate Bayesian Computation, tailored to handle the increasing parameter dimension in time. Their algorithm allows obtaining results timely enough to be usable in practical situation awareness. Their approach is in active use, informing the Norwegian crisis team about the regional epidemical situation in Norway.

Dr. De Blasio will comment on the practical use of modelling in the management of the pandemics by the Norwegian Institute of Public Health, presenting its impact and challenges.