Integrated Eigenspace Method with Spatial Proximity for Disease Cluster Detection
Keywords:
Eigenspace, nonparametric approach, cluster detection, spatial proximity, neighbourhoodAbstract
Spatio-temporal disease cluster detection is crucial for health organizations, especially in low and middle-income countries. These clusters provide useful information that can be used to improve surveillance systems for targeted diseases and allocate necessary resources for intervention effectively. Researchers recently proposed a new nonparametric Eigenspace-based method called Multi-EigenSpot that detects spatio-temporal disease clusters without restricting the quality, shape or distribution of data. In this algorithm, cluster is generally defined as a group of geographical regions that are spatially and temporally connected. However, this algorithm has a limitation that it does not explicitly consider spatial proximity when detecting the clusters. To overcome this limitation, this research introduces a retrospective approach for the Eigenspace-based algorithm by integrating spatial proximity for detection of disease cluster. The spatial proximity works by grouping each region with its spatially adjacent regions, rather than considering the disease clusters as individual cases. Highlight of this research is the spatial proximity where the temporal relationships are assumed to be constant in order to approximate borders for each location. The integrated eigenspace method was applied to monthly measles cases in Khyber Pakhtunkhwa, Pakistan in the year 2016. Results obtained were compared with Multi-EigenSpot. It was found that the integrated algorithm was able to identify more disease clusters (hotspots) which indicated improve and accurate prediction and shall be effective for public health intervention and control activities.