Conditional predictive inference for online surveillance of spatial disease incidence

Area of Science:

• Epidemiology
• Biostatistics
• Spatial Analysis

Background:

• Timely detection of disease clusters is crucial for public health.
• Multivariate surveillance techniques can improve localized outbreak detection compared to traditional methods.
• Spatial and temporal data availability enhances disease cluster surveillance capabilities.

Purpose of the Study:

• To develop statistical methodology for timely detection of incident disease clusters in space and time.
• To introduce a general Bayesian model-based surveillance technique for detecting localized increases in disease incidence.
• To address the multiple comparisons problem in disease cluster detection.

Main Methods:

• Development of the surveillance conditional predictive ordinate (SCPO) as a Bayesian surveillance technique.
• Incorporation of a common probability for false alarms to manage multiple comparisons.
• Utilizing Bayesian hierarchical Poisson models for performance evaluation.
• Conducting a simulation study to assess the technique's effectiveness.
• Application in a case study of salmonellosis in South Carolina.

Main Results:

• The proposed Bayesian method, SCPO, demonstrates potential for enhanced detection of localized disease clusters.
• Simulation studies indicate the technique's performance in identifying spatial-temporal disease patterns.
• The method effectively manages multiple comparisons, reducing false alarms.
• The salmonellosis case study illustrates practical application in real-world epidemiological surveillance.

Conclusions:

• The developed Bayesian methodology provides a robust framework for timely spatial-temporal disease cluster detection.
• The SCPO technique offers an improvement over traditional surveillance methods by leveraging spatial data.
• This approach is valuable for public health agencies in monitoring and responding to disease outbreaks.