From traffic to transmission: adapting an agent-based transport model to simulate the spreading of infectious diseases

The main idea of this thesis originated from the onset of the COVID-19 pandemic in 2020, driven by the need for computer models to assist policymakers in making decisions about mitigation strategies. The thesis presents an approach to simulate the spreading of infectious diseases, such as COVID-19, by integrating transport modeling with epidemiological methods. The work builds upon a MATSim transport model which captures movement patterns of a synthetic population based on mobile phone data. These patterns are used to identify contacts between synthetic persons. The contacts can be attributed to certain contexts as the MATSim model is activity-based, meaning that different activity types such as work, school or leisure are included in the model. Thus, this serves as a good starting point for constructing an agent-based epidemiological model, as the transport model inherently includes a synthetic population and potential infection pathways. Building on this foundation, the thesis presents a model that integrates the approaches from transport modeling with a mechanistic infection model and a disease progression model to simulate the spreading of infectious diseases. The model is validated against infection and hospital numbers in Berlin and Cologne. The model was developed during the COVID-19 pandemic to advise the German government. Major benefits of the model are its ability to quantify the contributions of different activity types to the overall infection dynamics and to assess the outcomes of different intervention scenarios. For example, the results show that applying moderate contact restrictions across all activity types is more effective than completely shutting down some activities while leaving others entirely unrestricted. During the pandemic, the model was continuously refined to incorporate evolving societal and po- litical questions. Examples include the wearing of masks, contact tracing, rapid testing and activity- specific restrictions. The emergence of multiple virus mutations, combined with the availability of vaccinations, introduced further levels of complexity to the model. To represent the influence of im- munity on infection dynamics, an extension was developed to calculate the protection against infection for each synthetic person, based on their individual immunization history. The model adopts a data-based approach, with the most important input being population activity participation over time derived from cell-based mobile phone data. This data source provides inherent information on how the population changed its mobility behavior in the course of the pandemic. The cell-based mobile phone data revealed, for example, that the German population noticeably reduced its activity participation well before the government implemented restrictions during the first COVID- 19 wave. One limitation of cell-based data is its inability to differentiate activity participation by activity types. It only allows for a distinction between “home” and “not at home”. In this context, a further contribution of this thesis is to propose an approach to compensate for this deficiency by using GPS-based mobile phone data in combination with OpenStreetMap to identify activity participation by activity types. Results show that GPS-based data is significantly more effective in distinguishing activity types, enabling, for instance, the differentiation of school-related from work or leisure activities.

Author dc:creator

• Müller, Sebastian Alexander

Advisor dc:contributor.advisor

• Nagel, Kai

Licence dc:rights.uri

• https://creativecommons.org/licenses/by/4.0/

Language dc:language.iso

en

Identifier URI

https://doi.org/10.14279/depositonce-23152

OAI identifier oai:identifier

oai:depositonce.tu-berlin.de:11303/24338