Analyzing 2000–2010 Childhood Age-Adjusted Cancer Rates in Florida: A Spatial Clustering Approach

Abstract

In this work, we apply a Bayesian hierarchical model that uses spatial clustering techniques to data from the Florida Association of Pediatric Tumor Programs (FAPTP) for the period 2000–2010. The goal is to determine whether there are statistically significant childhood cancer clusters at the Zip Code Tabulation Area (ZCTA) level of geography. The model provides estimates of the uncertainty associated with the clustering configurations, which is typically lacking in classical analyses of large datasets where a unique clustering representation can be insufficient. The model also allows covariate adjustment for known risk factors, bringing further relevant information, and it produces clusters that are spatially contiguous, enabling simple interpretation. The output clustering map is able to capture such patterns as the high-risk area that appear in the Southwest, Northeast, and Northwest Florida, which is consistent with the previous studies, but with finer details and deeper insight into year-specific features. New findings from the latest data, from 2008 to 2010, were also obtained and investigated. Our post-hoc validation of the clusters provides evidence for concluding that areas of elevated risk exist.