Distributed Multi-agent Architecture for Decision Support in Public Health Networks

One of the significant challenges that multi-user biosurveillance systems have is alarm management. Currently deployed syndromic surveillance systems [1–3] have a single user interface. However, different users have different objectives; the alarms that are important for one category of user are irrelevant to the objectives of another category of user. For example, a physician wants to identify disease on an individual-patient level, a county health authority is interested in identifying disease outbreak as early as possible within his local region, while an epidemiologist at the national level is interested in global situational awareness. The objective of a multi-agent decision support system is not only to recognize patterns of epidemiologically significant events but also to indicate their relevance to particular user groups’ objectives. Thus, instead of simply providing alerts of anomaly detections, the system architecture needs to provide analyzed information supporting multiple users’ decisions.