Enhancing spatial detection accuracy for syndromic surveillance with street level incidence data

Area of Science:

• Epidemiology
• Public Health Surveillance
• Geographic Information Systems (GIS)

Background:

• The Department of Defense Military Health System utilizes a global syndromic surveillance system, Electronic Surveillance System for Early Notification of Community-based Epidemics (ESSENCE).
• ESSENCE employs spatial and temporal algorithms to detect disease outbreaks from medical records at over 450 facilities worldwide.
• This study investigates enhancing ESSENCE's spatial detection by increasing data resolution from zip codes to street addresses.

Purpose of the Study:

• To evaluate the impact of geocoding medical incidence data to street addresses versus zip codes on the spatial accuracy of outbreak detection.
• To assess improvements in the localization and sizing of detected disease clusters using higher spatial resolution data.

Main Methods:

• Utilized Influenza-Like Illness (ILI) data from the 2008/2009 season for method development.
• Superimposed simulated disease clusters of varying sizes onto real ILI incidence data.
• Applied the spatial scan statistic to detect clusters and measured displacement from simulated locations.

Main Results:

• Street-level geocoding reduced cluster displacement by 65% for 2-5 mile radii and 31% for 10-mile radii compared to zip code methods.
• Street address data produced more accurate cluster size distributions, closely matching simulated radii.
• The traditional ESSENCE zip code method resulted in fragmented distributions and a high incidence of zero-radius or oversized clusters.

Conclusions:

• Geocoding to street addresses significantly improves the spatial accuracy of disease outbreak detection in terms of both location and cluster size.
• Despite improvements, only 73.5% of cases were successfully geocoded to street addresses, with zip codes still used for over a quarter of cases.
• Systematic improvements in collecting individual address information are crucial for advancing spatial detection accuracy in syndromic surveillance.