The Use of Geospatial Clustering in Analysing Health Risk Profile

• Main Authors: YEO, Sue-Mae, KAM, Tin Seong, THIA, Kai Xin, WU, Dan
• Format: text
• Language: English
• Published: Institutional Knowledge at Singapore Management University 2014
• Subjects: LISA; Spatial statistics; Population Health; MITB student; Computer Sciences; Databases and Information Systems; Geography; Medicine and Health Sciences
• Online Access: https://ink.library.smu.edu.sg/sis_research/2528; https://ink.library.smu.edu.sg/context/sis_research/article/3528/viewcontent/Geospatial_Health_Risk_2014_av.pdf
• Institution: Singapore Management University

Background & Hypothesis: The first law of geography states that “everything is related to everything else, but near things are more related than distant things”. This study aims to demonstrate how local indicator of spatial association (LISA) statistics are used to group patients with similar chronic diseases into natural clusters of hotspots found within northern Singapore by incorporating the proximity of their home locations explicitly. Methods: Anonymised chronic patient data collected from Khoo Teck Puat Hospital in 2013 were used for analyses. The data was mapped based on patients' residential addresses. A layer of hexagonal grid objects, each with a radius of 250 metres, was then generated and subsequently used to transform individual point data into area data. The local Moran statistical method was used to compute and test on each hexagonal grid object for significance by randomisation to identify clusters of hotspots. Results: Clusters of patients with chronic diseases were found in Nee Soon, Canberra and the intersection of Woodlands and Admiralty political divisions. For hypertension, clusters of patients aged 40 and above were found concentrated in Nee Soon political division. Discussion & Conclusion: The results showed that LISA statistics were more effective in delineating natural clusters as compared to conventional clustering method. The study also reported the statistical significance of each cluster. With these hotspots identified, healthcare intervention programmes can be customised according to the clusters found.