Generative analysis of dynamic sexual contact networks

 

GEORGE KAMPIS

COLLEGIUM BUDAPEST, LORAND EOTVOS UNIVERSITY

 

Managing and understanding HIV is an important challenge.The infection spreads via physical contacts, which naturally define a network. The dynamic nature of sexual contact networks has been the subject of a growing number of studies recently. In empirical data, differences between "cumulative" sexual contact networks recording partner relations over a longer period of time (e.g., over the entire lifetime of the persons involved) versus networks that capture a shorter snapshot of the contact network have been demonstrated. However, empirical data about sexual contact networks is very scarce and, understandably, typically available about aggregate properties only, such as degree distribution. Degree distributions, in turn, do not always contain the information necessary in order to detect or simulate the spreading of HIV. For example, degree distributions do not tend to indicate whether the network is connected, or tell about the size of the largest component etc., which are key factors for information/infection propagation. A theoretical understanding of dynamic contact networks is missing. Only a few works exist that directly address the problem of how individual level sexual behavior leads to different network structures and how this is related to the infection potential. We use agent-based modeling techniques to systematically generate different simple models of dynamic contact networks that yield both short term and cumulative distributions, similar to empirical data. We extend our studies to also study additional structural properties, including component distributions, etc. The main findings are that under a broad set of different natural conditions (such as different regimes of serial monogamy or concurrency) the resulting network is strongly disconnected and the typical time envelope is lognormal whereas the time local distribution can show scale-free behavior. We discuss how this bears on the spreading of the infection and on the significance of individual nodes for prevention.