Mapping the global potential transmission hotspots for severe fever with thrombocytopenia syndrome by machine learning methods
Mapping the global potential transmission hotspots for severe fever with thrombocytopenia syndrome by machine learning methods
复制标题
DOI：
10.1080/22221751.2020.1748521
复制
发表时间：
2020-01-01
影响因子：
13.2
通讯作者：
Fang, Li-Qun
中科院分区：
文献类型：
--
作者：
Miao, Dong;Dai, Ke;Fang, Li-Qun
重症发热伴血小板减少综合征(SFTS)是一种新发传染病，其传播范围不断扩大。然而，目前SFTS的传播已经扩展到亚洲以外的国家，确切的全球范围和风险模式仍然不清楚。在这里，我们建立了一个详尽的数据库，其中包括全球报告的人类短角血吸虫病例和可胜任媒介长角血蜱以及解释性环境变量，在此基础上，应用两种机器学习方法绘制了长角血吸虫的潜在地理范围和短角血吸虫的风险区域。确定了对全球天牛分布有贡献的10个预测因子，其相对贡献率为1%。在当代已知分布之外，我们预测在两个大陆，包括美国东北部、新西兰、澳大利亚的部分地区和几个太平洋岛屿，都会对长角毛虫有很高的接受度。确定了SFTS病例发生的8个关键驱动因素，包括海拔、预测存在长角毛虫的可能性、两个与温度有关的因素、两个与降水有关的因素、哺乳动物的丰富度和水体覆盖率。创建了全球模型预测的人类SFTS发生风险图，并验证了其在中国以外的三个报告病例的国家/地区中区分实际感染和未感染地区的有效性(预测中值概率为0.74vs.0.04，P<；0.001)。预测的高危地区(概率=50%)主要在中国中东部、朝鲜半岛大部分地区和日本南部以及新西兰北部。我们的发现强调了应该对潜在的SFT传播或入侵事件保持高度警惕的领域，因为它们对长角血吸虫的分布具有很高的接受性。
Severe fever with thrombocytopenia syndrome (SFTS) is an emerging infectious disease with increasing spread. Currently SFTS transmission has expanded beyond Asian countries, however, with definitive global extents and risk patterns remained obscure. Here we established an exhaustive database that included globally reported locations of human SFTS cases and the competent vector, Haemaphysalis longicornis (H. longicornis), as well as the explanatory environmental variables, based on which, the potential geographic range of H. longicornis and risk areas for SFTS were mapped by applying two machine learning methods. Ten predictors were identified contributing to global distribution for H. longicornis with relative contribution >= 1%. Outside contemporary known distribution, we predict high receptivity to H. longicornis across two continents, including northeastern USA, New Zealand, parts of Australia, and several Pacific islands. Eight key drivers of SFTS cases occurrence were identified, including elevation, predicted probability of H. longicornis presence, two temperature-related factors, two precipitation-related factors, the richness of mammals and percentage coverage of water bodies. The globally model-predicted risk map of human SFTS occurrence was created and validated effective for discriminating the actual affected and unaffected areas (median predictive probability 0.74 vs. 0.04, P < 0.001) in three countries with reported cases outside China. The high-risk areas (probability >= 50%) were predicted mainly in east-central China, most parts of the Korean peninsula and southern Japan, and northern New Zealand. Our findings highlight areas where an intensive vigilance for potential SFTS spread or invasion events should be advocated, owing to their high receptibility to H. longicornis distribution.