BEST PRACTICES - SPACE-TIME DISEASE SURVEILLANCE TOOLS

GeoSurveillance (version 1.1)

Features

Global score test
Local score test
Spacial M test

Parallel univariate cusums
Monitoring local cusum maxima
Linked table, graph and map displays

z-score transformations
Map animation

Comments

GeoSurveillance provides capabilities for testing three types of clustering: global clustering, local cluster detection, and focused clustering (i.e., around a pre-specified point location). These tests are spatial only when performed retrospectively, however GeoSurveillance can be used prospectively, using a cumulative sum approach combined with spatial clustering statistics. Linked displays allow for exploration of both individual and aggregate cusums and the map. GIS functions are limited to map display, and both text and shapefile data are supported as input formats.

Selected case studies

Rogerson & Yamada 2004 - Poisson CUSUM methods are outlined and applied to lower respiratory infection medical clinic visits in Boston over a three year period. Expected counts are modelled using logistic regression, and deviations from this expectation are monitored using the Poisson CUSUM approach. Adjustments for temporal trend and multiple testing are made. The analysis demonstrates that when monitoring deviations from a modelled expected value, much hinges on the ability to model the ‘in control' process accurately.

Meyer et al. 2008 - A syndromic surveillance system is described that integrated a number of syndromic data streams including nurse helpline telephone calls, laboratory test orders, and hotel staff absenteeism. Poisson CUSUM with time varying expected values was used to detect signals in the multivariate data. This system, deployed during the G8 Summit 2005 in Scotland, triggered 95 signals, of which 13 were investigated. The system successfully augmented traditional surveillance for early detection of natural and intentional disease outbreaks.

Surveillance Objective(s)

Cluster detection

User Expertise

Users of GeoSurveillance should be knowledgeable of spatial statistics before attempting analysis. There are parameters for both the retrospective spatial clustering tests (bandwidth) and the prospective CUSUM tests (bandwidth, slack, threshold). The statistical methodology of CUSUM has been developed in a series of papers by the authors, but limited information is available in the documentation itself.

Key Considerations

No batch mode for automated analyses
Examine the impact of bandwidth parameter by using a range of values
8 character limit on input shapefile names
Strict data structuring and field naming convention for prospective analyses

More Information

Website: http://www.acsu.buffalo.edu/~rogerson/geosurv.htm

Key Resource: Yamada, I, Rogerson P. & Lee, G. 2009. GeoSurveillance: a GIS-based system for the detection and monitoring of spatial clusters. Journal of Geographical Systems 11, no. 2: 155-173.

References

Rogerson, P. A., and I. Yamada. 2004. Approaches to syndromic surveillance when data consist of small regional counts. Morbidity and Mortality Weekly Report 53: 79-85.

Meyer, N., J. McMenamin, C. Robertson, M. Donaghy, G. Allardice, And D. Cooper. 2008. A Multi-Data Source Surveillance System to Detect a Bioterrorism Attack During the G8 Summit in Scotland. Epidemiology and Infection 136, no. 07: 876-885.