Choosing the best algorithm for event detection based on the intended application: A conceptual framework for...

Area of Science:

• Public Health
• Biostatistics
• Epidemiology

Background:

• Syndromic surveillance systems rely on early detection of epidemic signals.
• Numerous statistical methods exist for event detection, but a standardized classification is lacking.
• Selecting appropriate algorithms is challenging due to the absence of clear selection criteria linked to methods.

Purpose of the Study:

• To provide a practical framework for selecting statistical methods for event detection in syndromic surveillance.
• To guide practitioners in choosing algorithms based on specific application needs.
• To reduce technical barriers in developing and implementing syndromic surveillance systems.

Main Methods:

• Developed selection criteria by mapping algorithm assumptions and performance to time series characteristics.
• Considered expected epidemic types and surveillance system attributes.
• Focused on univariate and temporal methods, the most common in syndromic surveillance.

Main Results:

• A classification scheme for statistical methods used in syndromic surveillance event detection.
• Guidelines linking algorithm characteristics to data and system features.
• A practical approach to informed algorithm selection for practitioners.

Conclusions:

• The developed guidelines facilitate informed choices among popular event detection algorithms.
• This approach aids decision-makers, analysts, and researchers in syndromic surveillance.
• Enhances the development and implementation of effective public and animal health surveillance systems.