Analytic Biosurveillance Methods For Resource-Limited Settings

Public health surveillance faces many challenges in geographic regions lacking modern technology and infrastructure. This article addresses the role of analytic methods in such regions and evaluates temporal alerting algorithms using both authentic and simulated data sets. Evaluation analyses give the technical background for the statistical methods provided by the Johns Hopkins University Applied Physics Laboratory (APL) Suite for Automated Global Electronic bioSurveillance (SAGES), a collection of modular, open-source software tools to enable electronic surveillance in resource-limited settings. Included in the evaluation are only those statistical methods that are broadly applicable to multiple evolving-background time-series behaviors with limited data history. Multiple detection performance measures are defined, and a practical means of combining them is applied to recommend preferred alerting methods for common scenarios. Effective usage of these methods is discussed in the context of routine health-monitoring operations.