Analyzing and forecasting the Ebola incidence in North Kivu, the Democratic Republic of the Congo from 2018-19 in real time
Analyzing and forecasting the Ebola incidence in North Kivu, the Democratic Republic of the Congo from 2018-19 in real time
复制标题
DOI：
10.1016/j.epidem.2019.05.002
复制
发表时间：
2019-06-01
期刊：
影响因子：
3.8
通讯作者：
Nishiura, Hiroshi
中科院分区：
文献类型：
--
作者：
Akhmetzhanov, Andrei R.;Lee, Hyojung;Nishiura, Hiroshi
在埃博拉病毒病（EVD）暴发期间，实时分析和预测发病率受到病例报告，特别是报告延误的挑战。需要注意的是，实时统计的最新病例数可能被低估，而且在没有适当调整的情况下，有效再生产数（即单个原发性病例在给定时间点产生的平均继发病例数）也被低估。本研究旨在调整报告延迟，以适当估计刚果民主共和国北基伍省埃博拉病毒病的最新发病率，并从每周报告数据中获得短期预测。采用了一种半结构化建模方法，考虑了可能依赖于时间的报告延迟。平均报告延迟估计为11.6天（95% CI: 11.3, 11.9），标准差估计从2019年11月26日起从8.5-6.0天发生变化。临近预报通过考虑与时间相关的报告延迟成功实施：它主要包含95%置信区间内的未来观测值，但当报告的发生率随时间突然变化时，也会出现失败。预报也以与临近预报类似的方式进行，而我们对未来两周的有效再现数采用了外推方法。在最近一次观察时间之前几周的复制数的移动平均值优于其他推断。通过将建议的临近预报和预报集成到监测系统中，我们可以从实时（即连续）更新的“情况报告”中获得的信息大大改善。
During an Ebola virus disease (EVD) outbreak, the analysis and forecasting of the incidence in real time is challenged by reporting of cases, especially the reporting delay. It should be remembered that the latest count of cases is likely underestimated in real time, and moreover, the effective reproduction number, i.e. the average number of secondary cases generated by a single primary case at a given point in time, is also underestimated without proper adjustment. The present study aimed to adjust the reporting delay to appropriately estimate the latest incidence and obtain short-term forecasts from weekly reporting data of EVD in North Kivu, the Democratic Republic of the Congo (DRC). A semi-structured modeling approach was taken, accounting for reporting delay which can depend on time. The mean reporting delay was estimated at 11.6 days (95% CI: 11.3, 11.9) and the standard deviation was estimated to have changed from 26 November 2019 from 8.5-6.0 days. Nowcasting was successfully implemented by account for the time-dependent reporting delay: it mostly contained future observed values within the 95% confidence intervals, but there were failures when the reported incidence abruptly changed over time. Forecasting was also exercised in a similar manner to the nowcasting, while we imposed an extrapolation approach to the effective reproduction number for two future weeks. Moving average of the reproduction numbers for a few weeks prior the latest time of observation outperformed other extrapolations. The information that we can gain from real time (i.e. sequential) update of "situation report" can be considerably improved by integrating the proposed nowcasting and forecasting to the surveillance system.