Epidemic prediction and control in weighted networks.

Contact networks are often used in epidemiological studies to describe the patterns of interactions within a population. Often, such networks merely indicate which individuals interact, without giving any indication of the strength or intensity of interactions. Here, we use weighted networks, in which every connection has an associated weight, to explore the influence of heterogeneous contact strengths on the effectiveness of control measures. We show that, by using contact weights to evaluate an individual's influence on an epidemic, individual infection risk can be estimated and targeted interventions such as preventative vaccination can be applied effectively. We use a diary study of social mixing behaviour to indicate the patterns of contact weights displayed by a real population in a range of different contexts, including physical interactions; we use these data to show that considerations of link weight can in some cases lead to improved interventions in the case of infections that spread through close contact interactions. However, we also see that simpler measures, such as an individual's total number of social contacts or even just their number of contacts during a single day, can lead to great improvements on random vaccination. We therefore conclude that, for many infections, enhanced social contact data can be simply used to improve disease control but that it is not necessary to have full social mixing information in order to enhance interventions.