Regionalization of Youth and Adolescent Weight Metrics for the Continental United States Using Contiguity-Constrained Clustering and Partitioning

Abstract

Contemporary spatial data collection techniques, analyses, and presentations have created new opportunities for public health analyses that sometimes render existing administrative and statistical boundaries unsuitable. This article presents an applied algorithm, regionalization with dynamically constrained agglomerative clustering and partitioning (REDCAP), to create regions other than pre-defined regions. The regions created in the study were based on the weight of youth in the continental United States. The REDCAP algorithm incorporates a spatial contiguity restriction to create regions with the same characteristics and value. The regions created overcome the existing challenge in cartography in which administrative and statistical regions are often used in presenting results. The study generated 10- and 25-class regions that reflected high and low obesity prevalence among US youth without using existing county and state boundaries. The results revealed new insights about regions comprising counties identified as having high obesity prevalence. Some of the counties identified in the established regions interestingly have not been recorded as at risk for high obesity prevalence in previous studies. A crucial advantage of the approach is that it minimizes the bias contained in existing administrative and statistical regions, a challenge in cartography. Furthermore, the approach effectively creates regions based on a specific theme and objective function.