This project will create an agent-based simulation of Injection Drug Using networks to better understand how HIV does and does not spread over time. At the heart of this project is the well documented but as yet not understood phenomenon of sub-saturation stabilization of the HIV-1 virus among mature IDU communities - which stabilize despite the presence of new infection outbreaks in the network, and the ongoing practice of high risk behaviors. This simulation will take advantage of many years of research on HIV spreading and IDU networks. From these studies large amounts of topological, behavioral, and epidemiological data are available, such that more accurate network parameterization is now possible. In turn, this project will inform current empirical investigations by providing dynamic views of otherwise seeming static populations. Existing IDU network data can be mined for """"""""local attachment"""""""" dynamics, which can be used to produce a dynamic network, where individual actors or agents come and go over time, resort their connections, picking up new risk connections and leaving behind other, all according to the local rules drawn from existing networks. This data can be combined with realistic transmission probability curves to produce a dynamical network that combines an evolving topology with stochastic virus propagation. Within this simulation environment, various initial infection scenarios can be tested, and those cases where sub-saturation stabilization occurs can be isolated and analyzed for their relevant topological motifs. High numbers of runs from a range of initial starting points will produce a wide array of disease spreading histories across a range of network evolutes. Data mining and related analyses will be used to distinguish meaningful contributing network features. This strategy takes advantage of new design and analytical methods - network decompositions and motif mining techniques - to go beyond prior percolation simulations. These techniques will allow us to produce a dynamic model that invokes what Morris (2002) described as local rules and global properties: network actors for whom the network plays a structural role, rather than merely providing a structured environment. The public health risks associated with ongoing HIV transmission among IDU communities is immense, as individual injectors continue to maintain risk relationships with those outside the IDU community. If at-present unknown network topologies exhibit behavior that promotes the recurrence of new infections within contained areas, they can be seen as infection reservoirs, and thus a major health risk to both other injectors and the larger public. While simulation is not a substitute for empirical research, it can use the result of empirical research to go beyond the practical constraints of time and money associated with face to face data collection. This will be the first simulation on a scale with existing day urban drug communities (25,000 actors), and promises insight into time depths that empirical research cannot investigate. The team responsible for the completion of the project represents a multi-disciplinary collection of scholars at John Jay College, CUNY, with many years experienced in ethnography, social network theory, and graph theoretical research, all of whom have been active in HIV research in the New York area and with local community health partners.
Injecting drug users and their non-drug using partners continue to be at high risk of becoming infected with HIV and other blood borne pathogens through sharing injecting equipment, or by engaging in unprotected sex, with HIV+ persons - accounted for 13% of AIDS casese reported in the United States, with an additional 3% of cases among the heterosexual sex partners of IDUs. Yet HIV moves through IDU networks in unexpected ways, the dynamics of which are not well understood. This project will help researchers and policy makers better understand and prevent HIV propagation within IDU communities, and between IDU communities and the society in which they are embedded.
