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

Area of Science:

• Epidemiology
• Network Science
• Computational Biology

Background:

• Small populations like healthcare settings exhibit high contact heterogeneity impacting pathogen transmission.
• Empirical contact data is valuable but often limited in duration and subject to observation bias.

Purpose of the Study:

• To propose and validate an algorithm for stochastically reconstructing realistic temporal contact networks from limited individual contact data in healthcare settings.
• To address data gaps and observation bias in empirical contact network data.

Main Methods:

• Developed a novel algorithm using hourly contact rates, individual ward information, staff categories, and recurring contact frequency to generate full temporal networks.
• Incorporated an observation model to formalize recording bias and enable comparison between observed and reconstructed networks.
• Validated the algorithm using real-world data from a long-term care facility (LTCF) during the i-Bird study.

Main Results:

• The reconstructed networks demonstrated higher accuracy in reproducing network characteristics compared to random graphs.
• Successfully replicated key network properties such as assortativity by ward and hourly staff-patient contact patterns.
• Accurately reproduced the temporal correlation of contacts and consistently recreated unobserved contacts, generating complete networks for the LTCF.

Conclusions:

• The proposed algorithm effectively generates realistic temporal contact networks and reconstructs unobserved contacts from limited empirical data.
• This approach can be applied to various healthcare settings to create comprehensive contact networks for informing individual-based epidemic models.
• Enhances the utility of sparse contact data for understanding and modeling disease transmission dynamics.