Project Details
Projekt Print View

Utilization of spatially resolved data sources for an established agent-based model of Germany and its impact on predicted SARS-CoV-2 dynamics

Subject Area Epidemiology and Medical Biometry/Statistics
Data Management, Data-Intensive Systems, Computer Science Methods in Business Informatics
Term from 2021 to 2023
Project identifier Deutsche Forschungsgemeinschaft (DFG) - Project number 492390948
 
Final Report Year 2023

Final Report Abstract

Dynamic infection spread models are a key tool to support decision-makers in epidemically occurring infectious diseases. They synthesise the best currently available evidence with the aid of a mechanistic model to compare possible intervention strategies in realistic scenarios (so-called scenario modelling) or to obtain estimates of new infections to be expected within a defined period (so-called forecast modelling). Agent-based models offer the possibility of modelling infection events in a high degree of detail but require an equally detailed data basis. Due to the challenges involved, large agent-based models were rarely used for forecast modelling before and during the COVID-19 pandemic. In Space_impact, we used a global agent-based model developed for scenario modelling and extended it with an interface that allows the parameterisation of all spatially heterogeneous variables at the county level. The resulting modelling platform allows real-time daily forecasts of expected cases of new infections, hospitalisations, and intensive care patients over 7 to 30 days at the county level and allows decision-makers to interact directly via a newly developed dashboard. The developed model was extensively externally validated and demonstrated high regional prediction accuracy. However, incorporating regionally stratified data in real-time became increasingly difficult as, for example, mobility data became unavailable during the COVID-19 pandemic. As a result, we developed a dedicated model for deriving regional contact behaviour and mobility patterns based on contact studies conducted outside of Space_impact to provide contact and mobility behaviour in real-time, which is a core component of the higher-level modelling platform. Theoretical results of Space_impact were directly transferred into practical application in the BMBF projects OptimAgent and RespiNow, while several new methodological challenges were identified.

Publications

 
 

Additional Information

Textvergrößerung und Kontrastanpassung