Evaluating spatial surveillance: detection of known outbreaks in real data

Area of Science:

• Public Health
• Epidemiology
• Biostatistics

Background:

• Following anthrax and SARS outbreaks, interest in early illness detection systems has grown.
• Current systems monitor diverse data sources like healthcare visits and pharmacy sales.
• Limited evaluation exists for these new surveillance systems' performance.

Purpose of the Study:

• To propose and evaluate statistical methods for assessing novel disease surveillance systems.
• To determine if new systems signal true outbreaks more frequently than random chance.
• To compare the performance and timeliness of new systems against existing public health surveillance.

Main Methods:

• Developed three statistical methods to test the hypothesis that new systems do not signal outbreaks more than expected by chance.
• Methods vary in assumption restrictiveness, allowing flexibility in application.
• Analysis focuses on comparing signals from new systems with confirmed outbreaks from existing systems.

Main Results:

• The proposed methods can identify weaknesses in new surveillance systems.
• Confidence limits can be established for the agreement between new system signals and actual outbreaks.
• The methods assess if new systems detect outbreaks better than chance and if they provide earlier signals.

Conclusions:

• The proposed methods offer a framework for evaluating the effectiveness of new disease surveillance systems.
• These evaluations are essential for validating new tools for public health surveillance.
• The methods support decisions on adopting new systems as complements or alternatives to existing ones.