Detecting and Visualizing Space-Time Clusters of Low Birthweight Incidence, by Race, in Florida

• Activity Number: 186
• Type: Invited
• Date/Time: Tuesday, August 5, 2003 : 8:30 AM to 10:20 AM
• Sponsor: Section on Statistics & the Environment
• Author(s): Stephen Reader*+
• Companies: University of South Florida
• Keywords: low birthweight ; spatial statistics ; space time clusters ; Florida ; GIS

Abstract

There remains considerable debate about the role and significance of different risk factors associated with low birthweight incidence. Studies adopting an explicit geographic approach that might allude to environmental and other contextual factors beyond maternal risk factors are rare. This paper presents a method for detecting space-time clusters of low birthweight incidence when birth records are geocoded to point locations, and when the relative probabilities of low birthweight occurrence are prior-adjusted to reflect maternal risk factors. The analysis method uses logistic regression, spatial kernel density estimation, Monte-Carlo random-labeling methods, and GIS vector/raster functionality. Given the differences in low birthweight incidence between Blacks and Whites, the analysis is disaggregated by race. The analysis is conducted for the whole state of Florida using Vital Statistics records for 1992-1997. In addition to the identification of space-time clusters of low birthweight incidence, the paper also addresses the sensitivity of results to geocoding and the influence of spatial scale as controlled through the spatial kernel density methodology.