An agent-based model for estimating daily face-to-face contact networks in large urban systems

Abstract

Detailed contact data is important to model disease transmission in dense urban areas, as human mobility and social interactions significantly impact spread. However, linking mobility, activities, and social contacts in large cities is challenging. Current models often rely on contact surveys and overlook travel behaviors. Here we present a novel modeling framework for estimating large-scale, multi-setting contact networks by leveraging high-resolution synthetic activity-travel data. Our approach introduces a new mathematical formalism to construct multi-setting contact networks from spatiotemporal co-location patterns, enabling constraints based on key statistics (e.g., contact rates per setting, proportions of each contact type), incorporation of location types, and individual activity purposes. Efficient algorithms extract co-presence events, generating validated, individual-based contact networks, from which age-specific contact matrices are derived. The approach is tested using EQASIM, an open and reproducible activity-based travel demand model that relies on publicly available data for France’s Île-de-France region. We also evaluated the spatial effects of work-from-home policies on contact patterns by modifying individuals’ activity-travel diaries. Results show that multi-setting contact networks — representing 12 million individuals distributed across 1,714,920 unique locations — can be generated in minutes while accurately reproducing setting- and age-specific spatial contact patterns.