Distributed Information Fusion Networks as a Tool to Facilitate Regional Collaboration without Data Sharing

Introduction: We hypothesized that the situational awareness of epidemiological activities in neighboring jurisdictions will increase signals for events occurring within a given county. This assumption can be significant in large metropolitan areas where people regularly commute from one county to another. Also in counties where data is sparse, timely detection of events generally is a challenging problem. The Intelligent Decision Support Network (IDSN) was built to utilize available regional information and provide users with knowledge based on the regional situational awareness picture.

Method: The IDSN is composed of Bayesian Information Fusion Models (BIFM). Each BIFM supports one of the local health departments from a county in Northern Virginia. In addition to the data available to the local user, each BIFM integrates information received from other BIFMs representing neighboring jurisdictions. For example, within IDSN, the BIFM for one county receives that county’s data and estimates probability of the outbreak from BIFMs in neighboring counties.

 Results: Three neighboring counties in Northern Virginia were selected, one with sparse data and the others where data was dense. Two fusion models for the county with the sparse data were built. The first model received data from the local hospital and outpatient medical facility visits only. The second model was a BIFM.  In addition to the local data, the BIFM received estimated probability of influenza from the two neighboring counties’ BIFMs. Both models detected two seasonal influenza outbreaks. However, the BIFM detected both outbreaks earlier and showed higher probability of the influenza in the beginning of the outbreak as well as for all of the time periods when influenza cases were increasing.

 Conclusion: The results suggest that distributed architecture based on the exchange of information fusion results, instead of actual data sharing, can be effective.