A0331
Title: Regularized spatial and spatio-temporal cluster detection: Applications to breast cancer
Authors: Jun Zhu - University of Wisconsin - Madison (United States)
Ronald Gangnon - University of Wisconsin-Madison (United States)
Junho Lee - Louisiana State University (United States)
Maria Kamenetsky - University of Wisconsin-Madison (United States) [presenting]
Abstract: There are patterns in how people and disease group across space and time. These patterns are important to epidemiologists and health professionals because they may indicate elevated disease risk. In some cases, this high risk may be driven by external factors such as environmental exposures, infectious diseases, changes in lifestyle factors or other factors where a timely public health intervention may save lives. The detection of disease clusters has typically been approached as a large-scale multiple testing problem using a spatial and spatio-temporal scan statistic. Instead, spatial cluster detection has been re-examined as a high-dimensional variable selection problem using (quasi-)Poisson regression penalized by the least absolute shrinkage and selection operator (LASSO). Using sparse matrices, fast and efficient computation is made possible by exploiting the effects of potential clusters. Final models are selected based on (quasi-)information criteria, which allows us to smooth over the background rate and identify the selected breast cancer clusters. Data-driven simulation results demonstrate our approach for detecting single and multiple spatio-temporal clusters. Practical applications of the methods are illustrated using data on breast cancer incidence in Japan.
