The classical likelihood ratio spatial scan statistics has been widely used in spatial epidemiology for disease cluster detection. The question is whether the geographic incidence pattern is due to random fluctuations or the map reflects true underlying geographical variation due to etiologic risk factors. The hypothesis underlying the classic scan statistics assume that disease counts in different locations have independent Poisson distribution; unfortunately, outcomes in spatial units are often not independent of each other. Risk estimates of areas that are close to each other will tend to be positively correlated as they share a number of spatially varying characteristics. Ignoring the overdispersion caused by spatial autocorrelation leads to incorrect results. To overcome this difficulty, we propose a model-based approach adjusting for area-specific fixed-effects measuring potential effect modifiers, and for large-scale geographical variation of etiologic factors that vary continuously in space and are not expressly present within the model. We apply our methodology to the spatial distribution of lung cancer male mortality occurred in the province of Lecce, Italy, during the period 1992-2001. © 2009 Springer Berlin Heidelberg.
CITATION STYLE
Bilancia, M., Montrone, S., & Perchinunno, P. (2009). A model-based scan statistics for detecting geographical clustering of disease. In Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) (Vol. 5592 LNCS, pp. 353–368). https://doi.org/10.1007/978-3-642-02454-2_25
Mendeley helps you to discover research relevant for your work.