Comparison of filtering methods for the modeling and retrospective forecasting of influenza epidemics.
Comparison of filtering methods for the modeling and retrospective forecasting of influenza epidemics.
复制标题
DOI：
10.1371/journal.pcbi.1003583
复制
发表时间：
2014-04
影响因子：
4.3
通讯作者：
Shaman J
中科院分区：
文献类型：
--
作者：
Yang W;Karspeck A;Shaman J
多种滤波方法可以实现系统状态变量的递归估计和模型参数的推断。这些方法已应用于一系列学科和环境，包括工程设计和预测，并且在过去二十年中已应用于传染病流行病学。对于任何感兴趣的系统，理想的滤波器取决于所应用模型的非线性和复杂性、所夹带观测值的质量和丰富性以及最终应用（例如预测、参数估计等）。在这里，我们比较了六种最先进的过滤方法在建模和预测流感活动时的性能。三个粒子滤波器（具有重采样和正则化功能的基本粒子滤波器 (PF)、通过迭代滤波的最大似然估计 (MIF) 和粒子马尔可夫链蒙特卡罗 (pMCMC)）以及三个集成滤波器（集成卡尔曼滤波器 (EnKF)、集成调整卡尔曼滤波器 (EAKF) 和秩直方图滤波器 (RHF)）与湿度强制滤波器结合使用易感-感染-恢复-易感（SIRS）模型和每周流感发病率估计。该建模框架首先使用合成流感流行数据进行验证，然后用于拟合和回顾性预测 2003 年至 2012 年期间美国 115 个城市七次流感流行的历史发病时间序列。结果表明，当使用 SIRS 模型时，集合过滤器和基本 PF 更能够忠实地重现历史流感发病时间序列，而 MIF 和 pMCMC 对于多模式爆发的表现则不佳。对于流感活动最高的一周的预测，六个模型过滤框架的准确性具有可比性；三个粒子过滤器在预测未来 1-5 周的峰值方面表现稍好；集成滤波器可以更准确地预测过去的峰值。流感是美国的一个重大公共卫生负担，每年导致 3,000 至 49,000 人死亡。如果流感预测可靠，将为公共卫生官员提供宝贵的预警，有助于减轻这种疾病的负担。例如，包括疫苗和抗病毒药物在内的医疗资源可以在流感发病高峰之前尽早分发到有需要的地区。最近统计过滤方法在流行病学模型中的应用表明，准确可靠的流感预测是可能的；然而，存在许多过滤方法，并且任何过滤器的性能可能取决于应用。在这里，我们使用单一流行病学建模框架来测试六种最先进的流感建模和预测过滤器的性能。其中三种过滤器是粒子过滤器，常用于科学、工程和经济学科；其他三个滤波器是集合滤波器，经常用于地球物理学科，例如数值天气预报。我们使用这六个过滤器中的每一个对 2003 年至 2012 年美国 115 个城市的季节性流感活动进行回顾性建模和预测。我们报告了这六个过滤器的性能并讨论了改进实时流感预测的潜在策略。
A variety of filtering methods enable the recursive estimation of system state variables and inference of model parameters. These methods have found application in a range of disciplines and settings, including engineering design and forecasting, and, over the last two decades, have been applied to infectious disease epidemiology. For any system of interest, the ideal filter depends on the nonlinearity and complexity of the model to which it is applied, the quality and abundance of observations being entrained, and the ultimate application (e.g. forecast, parameter estimation, etc.). Here, we compare the performance of six state-of-the-art filter methods when used to model and forecast influenza activity. Three particle filters—a basic particle filter (PF) with resampling and regularization, maximum likelihood estimation via iterated filtering (MIF), and particle Markov chain Monte Carlo (pMCMC)—and three ensemble filters—the ensemble Kalman filter (EnKF), the ensemble adjustment Kalman filter (EAKF), and the rank histogram filter (RHF)—were used in conjunction with a humidity-forced susceptible-infectious-recovered-susceptible (SIRS) model and weekly estimates of influenza incidence. The modeling frameworks, first validated with synthetic influenza epidemic data, were then applied to fit and retrospectively forecast the historical incidence time series of seven influenza epidemics during 2003–2012, for 115 cities in the United States. Results suggest that when using the SIRS model the ensemble filters and the basic PF are more capable of faithfully recreating historical influenza incidence time series, while the MIF and pMCMC do not perform as well for multimodal outbreaks. For forecast of the week with the highest influenza activity, the accuracies of the six model-filter frameworks are comparable; the three particle filters perform slightly better predicting peaks 1–5 weeks in the future; the ensemble filters are more accurate predicting peaks in the past. Influenza, or the flu, is a significant public health burden in the U.S. that annually causes between 3,000 and 49,000 deaths. Predictions of influenza, if reliable, would provide public health officials valuable advanced warning that could aid efforts to reduce the burden of this disease. For instance, medical resources, including vaccines and antiviral drugs, can be distributed to areas in need well in advance of peak influenza incidence. Recent applications of statistical filtering methods to epidemiological models have shown that accurate and reliable influenza forecast is possible; however, many filtering methods exist, and the performance of any filter may be application dependent. Here we use a single epidemiological modeling framework to test the performance of six state-of-the-art filters for modeling and forecasting influenza. Three of the filters are particle filters, commonly used in scientific, engineering, and economic disciplines; the other three filters are ensemble filters, frequently used in geophysical disciplines, such as numerical weather prediction. We use each of the six filters to retrospectively model and forecast seasonal influenza activity during 2003–2012 for 115 cities in the U.S. We report the performance of the six filters and discuss potential strategies for improving real-time influenza prediction.
登录
查看更多内容
影响因子：
16.6
作者：
Shaman, Jeffrey;Karspeck, Alicia;Yang, Wan;Tamerius, James;Lipsitch, Marc
通讯作者：
Lipsitch, Marc
影响因子：
3.7
作者：
通讯作者：
--
影响因子：
4.5
作者：
Ionides, Edward L.;Bhadra, Anindya;King, Aaron
通讯作者：
King, Aaron
DOI：
10.1073/pnas.0806852106
发表时间：
2009-03-03
影响因子：
11.1
作者：
Shaman, Jeffrey;Kohn, Melvin
通讯作者：
Kohn, Melvin
影响因子：
4.9
作者：
Karspeck, Alicia R.;Anderson, Jeffrey L.
通讯作者：
Anderson, Jeffrey L.