Shaweno Debebe, Trauer James M, Denholm Justin T, McBryde Emma S
Department of Medicine, University of Melbourne, Melbourne, Victoria, Australia.
Victorian Tuberculosis Program at the Peter Doherty Institute for Infection and Immunity, Melbourne, Victoria, Australia.
R Soc Open Sci. 2018 Sep 19;5(9):180887. doi: 10.1098/rsos.180887. eCollection 2018 Sep.
Geospatial tuberculosis (TB) hotspots are hubs of TB transmission both within and across community groups. We aimed to quantify the extent to which these hotspots account for the spatial spread of TB in a high-burden setting. We developed spatially coupled models to quantify the spread of TB from geographical hotspots to distant regions in rural Ethiopia. The population was divided into three 'patches' based on their proximity to transmission hotspots, namely hotspots, adjacent regions and remote regions. The models were fitted to 5-year notification data aggregated by the metapopulation structure. Model fitting was achieved with a Metropolis-Hastings algorithm using a Poisson likelihood to compare model-estimated notification rate with observed notification rates. A cross-coupled metapopulation model with assortative mixing by region closely fit to notification data as assessed by the deviance information criterion. We estimated 45 hotspot-to-adjacent regions transmission events and 2 hotspot-to-remote regions transmission events occurred for every 1000 hotspot-to-hotspot transmission events. Although the degree of spatial coupling was weak, the proportion of infections in the adjacent region that resulted from mixing with hotspots was high due to the high prevalence of TB cases in a hotspot region, with approximately 75% of infections attributable to hotspot contact. Our results suggest that the role of hotspots in the geospatial spread of TB in rural Ethiopia is limited, implying that TB transmission is primarily locally driven.
地理空间结核病热点地区是社区群体内部和不同群体之间结核病传播的中心。我们旨在量化在高负担环境下这些热点地区对结核病空间传播的影响程度。我们开发了空间耦合模型，以量化埃塞俄比亚农村地区结核病从地理热点地区向遥远地区的传播情况。根据与传播热点的距离，将人群分为三个“斑块”，即热点地区、相邻地区和偏远地区。这些模型与根据集合种群结构汇总的5年通报数据进行拟合。使用Metropolis-Hastings算法，通过泊松似然将模型估计的通报率与观察到的通报率进行比较，从而实现模型拟合。根据偏差信息准则评估，一个按地区进行分类混合的交叉耦合集合种群模型与通报数据拟合得很好。我们估计，每1000次热点地区之间的传播事件中，有45次是从热点地区传播到相邻地区的事件，有2次是从热点地区传播到偏远地区的事件。尽管空间耦合程度较弱，但由于热点地区结核病病例的高流行率，相邻地区因与热点地区混合而导致的感染比例较高，约75%的感染可归因于与热点地区的接触。我们的结果表明，热点地区在埃塞俄比亚农村地区结核病地理空间传播中的作用有限，这意味着结核病传播主要是由当地因素驱动的。
