Related Experiment Videos
Measuring outbreak-detection performance by using controlled feature set simulations
Kenneth D Mandl1, B Reis, C Cassa
1Division of Emergency Medicine, Children's Hospital Boston, 300 Longwood Avenue, Boston, MA 02115, USA. Kenneth.Mandl@childrens.harvard.edu
MMWR Supplements
|February 18, 2005
Summary
This study introduces a flexible method using simulated outbreaks to evaluate public health surveillance systems. This approach enhances the detection of disease outbreaks by testing algorithms with realistic data.
Area of Science:
- Public Health Surveillance
- Epidemiology
- Biostatistics
Background:
- Evaluating outbreak detection performance requires benchmarking against real-world data.
- Limited real-world data for rare events like bioterrorism necessitates simulation.
- Semisynthetic data combines authentic disease baselines with simulated outbreaks for realistic noise and signal.
Purpose of the Study:
- To define a flexible approach for evaluating public health surveillance systems.
- To demonstrate the use of this approach in early outbreak detection.
- To provide a framework for optimizing surveillance system performance.
Main Methods:
- Describing stages of outbreak detection.
- Creating benchmark datasets using semisynthetic data with controlled simulated outbreaks.
- Defining outbreak signal parameters (size, shape, duration) and proposing detection metrics.
Main Results:
- Demonstrated flexibility of controlled feature set simulation for evaluating sensitivity and specificity.
- Optimized detection algorithm attributes, syndrome groupings, and data integration strategies.
- Validated the effectiveness of the semisynthetic data approach.
Conclusions:
- Semisynthetic data with controlled simulated outbreaks is valuable for benchmarking syndromic surveillance systems.
- This method allows for rigorous evaluation of detection performance.
- Facilitates the optimization of public health surveillance strategies.