Identifying hotspots of human anthrax transmission using three local clustering techniques

摘要

This study compared three local cluster detection methods to identify local hotspots of human cutaneous anthrax (HCA) transmission in the country of Georgia where cases have been steadily increasing since the dissolution of the Soviet Union. Recent reports have indicated that the disease has reached historical levels in 2012 highlighting the need for better informed policy recommendations and targeted control measures. The purpose of this paper was to identify spatial clusters of HCA to aid in the implementation of targeted public health interventions. At the same time, we compared the utility of different statistical tests in identifying hotspots. We used the Getis-Ord (G(i)* (d)), a multidirectional optimal ecotope-based algorithm (AMOEBA) - a cluster morphology statistic, and the spatial scan statistic in SaTScan (TM). Data on HCA cases from 2000 to 2012 at the community level were aggregated to an 8 x 8 km grid surface and population data from the Global Rural and Urban Mapping Project (GRUMP) were used to calculate local incidence. In general, there was agreement between tests in the locations of HCA hotspots. Significant local clusters of high HCA incidence were identified in the southern, eastern and western regions of Georgia. The G(i)*(d) and spatial scan statistics appeared more sensitive but less specific than the AMOEBA algorithm. The scan statistic identified larger geographic areas as hotspots of transmission. In general, the spatial scan statistic and G(i)* (d) performed well for spatial clusters with lower incidence rates, whereas AMOEBA was well suited for defining local spatial clusters of higher HCA incidence. In resource constrained areas, efficient allocation of public health interventions is crucial. Our findings identified hotspots of HCA that can be used to target public health interventions such as livestock vaccination and training on proper outbreak management. This paper illustrates the benefits of evaluating statistical approaches for defining disease hotspots and highlights differences in these clustering approaches applicable beyond public health studies. (C) 2015 The Authors. Published by Elsevier Ltd.