Detecting space-time disease clusters with arbitrary shapes and sizes using a co-clustering approach

Area of Science:

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

Background:

• Effective disease surveillance requires identifying space-time disease clusters.
• Traditional methods using fixed shapes (circles, ellipses) struggle with irregularly shaped disease occurrence patterns.
• Existing algorithms are often impractical for detecting clusters in non-uniform spatial distributions.

Purpose of the Study:

• To propose a novel co-clustering algorithm for detecting space-time disease clusters.
• To overcome the limitations of shape and size restrictions in existing cluster detection methods.
• To analyze malaria occurrences in Khyber Pakhtunkhwa Province, Pakistan, using spatio-temporal data.

Main Methods:

• Development of a co-clustering strategy for detecting prospective and retrospective space-time disease clusters.
• Tracking changes in space-time occurrence structure rather than exhaustive spatial searching.
• Application to annual and monthly malaria data from Khyber Pakhtunkhwa, Pakistan (2012-2016).
• Visualization of results using heat maps.

Main Results:

• The algorithm successfully identified potential space-time clusters in malaria data.
• Annual data analysis revealed significant hotspots in three sub-regions during 2013-2014.
• Monthly data analysis indicated recurring hotspots from July to October, demonstrating a strong seasonal pattern.
• The method proved effective for irregularly shaped disease clusters.

Conclusions:

• The proposed co-clustering algorithm offers a flexible and effective approach for space-time disease cluster detection.
• This method enhances disease surveillance capabilities by identifying non-geometrically shaped disease hotspots.
• Findings highlight the importance of seasonal monitoring for malaria prevention in the studied region.