AGENT-BASED APPROACH TO CONTACT PATTERNS MODELLING

Sunday, October 23, 2011
Grand Ballroom AB (Hyatt Regency Chicago)
Poster Board # 59
(MET) Quantitative Methods and Theoretical Developments

Marija Zivkovic Gojovic, PhD1, Beate Sander, RN, MBA, MEcDev1, Ashleigh Tuite2 and Natasha Crowcroft, PhD1, (1)Ontario Agency for Health Protection and Promotion, Toronto, ON, Canada, (2)University of Toronto, Toronto, Ontario, Toronto, ON, Canada

Purpose: The study is designed to develop a computational tool in order to establish and evaluate contact patterns relevant to infectious disease spread and thus to increase the application of agent-based technology in decision making process.

Method: Rapid development of computational capabilities in recent years has lead to the significant increase in the application of agent based modeling technology across various fields, including modeling of infectious disease spread. Defining the agent as an individual and the interactions between agents as social interactions the agent-based approach is often used to create an artificial society as a replica to the real world. This “artificial world” is further used as an experimenting tool to evaluate the wide range of different hypothesis, creating the “virtual laboratory” where experimental hazard is reduced to zero. The use of this tool is especially important when the real experiment can be categorized as costly, impractical or unethical.We developed the computational tool used to create an artificial society of a typical Canadian city and to establish major contact patterns between individuals based entirely on Statistics Canada Census Tract data.

Result: Contact patterns projected by the model showed a high level of similarity to the contact patterns observed in recent empirical studies: strong diagonal assertiveness in the age-based contact matrix which coincides with the age-preferred mixing patterns, existence of sub-diagonals related to children-parent relationships and overall reduction in contact intensity on weekends specifically within the same age groups (main diagonal). Sensitivity analysis of the model parameters additionally pointed that changes in the intensity and duration of contacts can highly influence the structure of the projected contacts patterns.

Conclusion: Comparison of the model outcomes to the empirically collected data showed that it is possible to use the agent based approach in order to create the realistic contact patterns. However, in order to use the agent-based model as an accurate tool in decision-making process, it is important to first determine the plausible ranges of values for sensitive parameters for a specific infectious agent.