Bayesian information fusion networks for biosurveillance applications.

Area of Science:

• Public Health
• Epidemiology
• Biostatistics

Background:

• Disease surveillance systems are crucial for public health.
• Existing systems may benefit from advanced data fusion techniques.
• Bayesian methods offer a robust framework for integrating diverse data streams.

Purpose of the Study:

• To develop and evaluate novel information fusion algorithms for enhanced disease surveillance.
• To incorporate Bayesian decision support capabilities into a detection system.
• To improve the specificity of outbreak detection.

Main Methods:

• A detection system was developed using chief complaints from emergency department visits.
• International Classification of Diseases Revision 9 (ICD-9) codes from outpatient records were utilized.
• Influenza surveillance data from health departments in the National Capital Region (NCR) were integrated.
• Bayesian Networks were employed to fuse data from multiple sources.
• Data anomalies and time offsets between data streams were analyzed.

Main Results:

• The developed system demonstrated increased specificity in identifying influenza-like events.
• Performance was compared against temporal anomaly detection algorithms used by NCR health departments.
• The Bayesian Network effectively fused data from diverse sources.

Conclusions:

• The proposed information fusion algorithms enhance disease surveillance systems.
• Bayesian decision support capabilities improve the accuracy of outbreak detection.
• Further research into data source correlations is recommended for more efficient data fusion.