A1263
Title: Spatial detection of adjacent hotspot clusters in disease mapping
Authors: Kunihiko Takahashi - Institute of Science Tokyo (Japan) [presenting]
Hideyasu Shimadzu - Kitasato University (Japan)
Abstract: In spatial epidemiology, statistical tests are often employed to detect regional disease clusters, particularly to evaluate whether disease risk is significantly elevated compared to surrounding areas. A central method is the cluster detection test (CDT), which identifies non-random spatial distributions and highlights high-prevalence regions without prior assumptions. Among CDT approaches, scan statistics based on maximum likelihood ratios, such as Kulldorff's circular scan and Tango and Takahashi's flexibly shaped scan, have been widely applied. More recent developments enable the simultaneous detection of multiple clusters by integrating generalized linear models with information criteria. However, conventional scan-based methods often assume uniform risk within a single cluster, leading to the erroneous merging of adjacent hotspots with different risk levels. To address this limitation, a new scan-based procedure is proposed, incorporating Cochran's Q-statistic to evaluate heterogeneity within clusters. This extended framework offers the accurate identification of adjacent hotspots as distinct clusters. Through real-world applications, the improved performance of the proposed method compared with existing scan-based tests is demonstrated.
