Spatiotemporal Analysis of the 2014 Ebola Epidemic in West Africa

In 2014–2016, Guinea, Sierra Leone and Liberia in West Africa experienced the largest and longest Ebola epidemic since the discovery of the virus in 1976. During the epidemic, incidence data were collected and published at increasing resolution. To monitor the epidemic as it spread within and between districts, we develop an analysis method that exploits the full spatiotemporal resolution of the data by combining a local model for time-varying effective reproduction numbers with a gravity-type model for spatial dispersion of the infection. We test this method in simulations and apply it to the weekly incidences of confirmed and probable cases per district up to June 2015, as reported by the World Health Organization. Our results indicate that, of the newly infected cases, only a small percentage, between 4% and 10%, migrates to another district, and a minority of these migrants, between 0% and 23%, leave their country. The epidemics in the three countries are found to be similar in estimated effective reproduction numbers, and in the probability of importing infection into a district. The countries might have played different roles in cross-border transmissions, although a sensitivity analysis suggests that this could also be related to underreporting. The spatiotemporal analysis method can exploit available longitudinal incidence data at different geographical locations to monitor local epidemics, determine the extent of spatial spread, reveal the contribution of local and imported cases, and identify sources of introductions in uninfected areas. With good quality data on incidence, this data-driven method can help to effectively control emerging infections.

2014—2016年，西非几内亚、塞拉利昂与利比里亚遭遇了自1976年埃博拉病毒被发现以来规模最大、持续时间最长的埃博拉疫情流行。疫情期间，研究人员以不断提升的时空分辨率收集并发布了发病例数数据。为监测埃博拉病毒在各行政区内部及跨行政区的传播态势，本研究开发了一种分析方法：该方法将用于计算时变有效再生数的局域模型与描述感染空间扩散的重力型模型相结合，充分利用了数据的全时空分辨率。我们通过仿真实验对该方法进行了验证，并将其应用于世界卫生组织（World Health Organization，WHO）报告的、截至2015年6月的各行政区确诊与临床诊断病例周度发病例数数据。研究结果显示，在新增感染病例中，仅有4%至10%的比例会迁移至其他行政区；而这类跨区迁移的病例中，仅0%至23%会离开其所在国家。分析表明，三国的疫情在估算得到的有效再生数以及行政区感染输入概率方面均较为相似。尽管敏感性分析提示该差异可能与疫情漏报有关，但三国在跨境传播环节中或扮演了不同的角色。该时空分析方法可利用不同地理位置的纵向发病例数数据，实现本地疫情监测、空间扩散范围判定、本地与输入病例贡献解析，以及未感染区域感染引入来源识别。依托高质量的发病数据，这一数据驱动方法可有效助力新兴感染性疾病的防控工作。