ESSENCE: Biosurveillance Detection Strategies with DARPA Evaluation Data

Multivariate and multiple univariate statistical process control methods were applied to over two years of multisource, multiregion data from a DARPA system-evaluation exercise. The research by Howard Burkom of Johns Hopkins University Applied Physics Laboratory demonstrates a substantial alerting-timeliness advantage for biosurveillance systems. The best methods achieved a 1-day median time before alert for a set of test outbreaks identified by epidemiologists.

Use Cases

• Evaluating spatial and temporal data-aggregation strategies for syndromic surveillance based on the described multisource environment.
• Optimizing algorithm performance for timely outbreak detection based on the comparative parametric analyses of temporal aggregation.
• Testing the sensitivity and timeliness of detection methods at practical false-alert rates based on the empirical thresholds calculated in the study.

Strengths

• Analysis is based on over two years of data records from a complex, authentic, multisource, multiregion environment.
• Outbreaks in the data were identified by a team of epidemiologists, providing a validated ground truth for evaluation.
• The best detection methods achieved a 1-day median time before alert over the set of test outbreaks.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.
• Last update date is unknown; freshness unverified.

Provenance

Source

Johns Hopkins University Applied Physics Laboratory

Collection Method

Data collection from a system-evaluation exercise for the Defense Advanced Research Project Agency (DARPA).

Time Range

Over 2 years (specific dates unknown)

Geography

Multiregion (specific regions unknown)