Abstract / Summary
Abstract Background With the increasing availability of high-resolution rate data such as age-specific mortality and disease incidence rates, it has become feasible to perform cluster detection stratified by age group. However, the accurate identification of spatial clusters in small populations remains challenging due to population sparsity and low event rates. To address this challenge, this study proposes a smoothing-based analytical framework to improve cluster detection performance in SaTScan under sparse and unstable conditions. Methods Three smoothing approaches, the Whittaker–Henderson (WH), Partial Standardized Mortality Ratio (PSMR), and Besag–York–Mollié (BYM2) methods, were evaluated through simulation studies across varying population sizes and age groups. Results The results indicate that all methods substantially improve the detection power of SaTScan compared to unsmoothed data, while WH achieves a better balance between statistical power and false detection rates. When applied to Taiwan’s 2021 female cancer mortality data, the WH method identified a significant hotspot among women aged 70 to 74 in southern Taiwan, primarily associated with elevated lung and liver cancer mortality. Conclusions The findings suggest that WH smoothing provides a practical and statistically reliable solution for stabilizing mortality estimates and identifying meaningful spatial clusters in small-population contexts. This framework offers valuable methodological and applied insights for spatial epidemiology and small-area health surveillance.