An algorithm to build synthetic temporal contact networks based on close-proximity interactions data

• DOI: 10.1371/journal.pcbi.1012227
• Publication Date: 2024-06-13T17:58:51Z
• Authors: Audrey Duval; Quentin J. Leclerc; Didier Guillemot; Laura Temime; Lulla Opatowski
• Keywords: 570; [INFO.INFO-MA]Computer Science [cs]/Multiagent Systems [cs.MA]; QH301-705.5; 610; [SDV.SPEE]Life Sciences [q-bio]/Santé publique et épidémiologie; Biology (General); [INFO.INFO-MO]Computer Science [cs]/Modeling and Simulation; 3. Good health; Research Article

Small populations (e.g., hospitals, schools or workplaces) are characterised by high contact heterogeneity and stochasticity affecting pathogen transmission dynamics. Empirical individual data provide unprecedented information to characterize such increasingly available, but usually collected over a limited period, can suffer from observation bias. We propose an algorithm stochastically reconstruct realistic temporal networks in healthcare settings (HCS) test this approach using real previously long-term care facility (LTCF). Our generates full recorded close-proximity interactions, hourly inter-individual rates on individuals’ wards, the categories of staff involved contacts, frequency recurring contacts. It also provides augmentation reconstructing contacts for days when some individuals present HCS without having empirical data. Recording bias is formalized through model, allow direct comparison between augmented observed networks. validate our during i-Bird study, compare reconstructed The was substantially more accurate reproduce network characteristics than random graphs. reproduced well assortativity ward (first–third quartiles observed: 0.54–0.64; synthetic: 0.52–0.64) patient patterns. Importantly, correlation (0.39–0.50 vs 0.37–0.44), indicating that could recreate structure. consistently recreated unobserved generate LTCF. To conclude, we summary statistics computed individual-level interaction This be applied extended other data, subsequently inform individual-based epidemic models.