Temporal network structure and its implication for disease dynamics and control

Background/Question/Methods

Contacts between individuals form the infrastructure over which diseases spread. Such contact patterns are far from random—there are correlations both in the network of who has been in contact with whom, and when these contacts happen. These structures affect the dynamics of disease spreading, but can also be exploited in preventive action such as immunization programs. In this work, we started from empirical, human contact structures—datasets from e.g. the proximity of patients in hospitals and Internet-mediated prostitution—to investigate temporal network structures and evaluate their effects on disease spreading. Our main tool was to run standard disease spreading models, like the SIR, on the contact structures (i.e., so that a disease can spread between two individuals at the times they are in contact). We used various types of randomizations and standard contact models as null-models to identify the effects of the temporal structure. We also investigated counter measures, like targeted vaccination, that were designed to exploit such local temporal structures.

Results/Conclusions

We found that the bursty dynamics of humans tend to speed up the early phase of an epidemic outbreak (compared to contact-models where contacts happen at a random times). Empirical data often has fat-tailed distributions of contact rates; something that in traditional network-epidemiological models is known to blur epidemic thresholds. Our study showed that empirical temporal structures tend to reinstate well-defined epidemic thresholds, in transmission probability but not in disease duration. Furthermore, we found that a particularly efficient way of stopping an outbreak in many types of human data is to vaccinate a fraction of the population by first calling people at random, then ask that person about their most recent contact, and vaccinating this contact. In such a way, one is able to find individuals that are currently in a period of more intense social (and potential disease-spreading) activity, and such individuals are more important in the disease spreading, so they are more important to vaccinate.