Forecasting influenza epidemics using a mechanistic epidemic model

摘要/摘要 这个项目将填补流感流行病学的一个基本知识空白，即缺乏 病毒抗原漂移与人类对流感易感性之间的量化关系。然后我们会申请 这一新知识有助于提高流感预报的准确性和及时性。抗原漂移是指 流感病毒表面蛋白的逐渐变化，这使得新的病毒株能够逃脱获得性 免疫，并使以前感染过流感的人再次感染。抗原性制图可以 量化流感病毒毒株之间的抗原漂移的大小(即抗原距离)。到目前为止， 然而，抗原距离和感染易感性之间的关系还没有被量化。 人类流感。 我们将使用流感传播和免疫的机制模型来估计这种联系 增加抗原距离和增加感染流感的易感性之间的关系。为此，我们将 利用独特的数据资源：自2011/12年流感以来进行的积极流感监测 季节通过美国流感疫苗效力网络。这些数据包括基于人口的估计 流感发病率，按病毒亚型/谱系以及抗原性和基因特征分层 在美国三个地理位置不同的州，传播流感病毒。这些数据还包括流感 目标人群的疫苗覆盖率。我们将把我们的机制流感模型应用于这些数据，并 量化漂移/磁化率关联。 然后，我们将应用这些发现来改进对季节性流感流行的预测。两种不同的 目前正在采取方法进行流感预测。短期预报使用近乎实时的监测数据 预测流感病例高峰期的时间和强度，提前几周。长期的 预测使用不同流感病毒株的相对流行率数据来预测哪些病毒株将占主导地位。 即将到来的一季。目前，无论是短期还是长期预测方法都没有有效地利用数据 由于接种疫苗或流感毒株的先前传播而对流感具有预先存在的免疫力。在量化之后 对于漂移/易感性关联，我们将测试我们的流感模型的预测能力。我们假设 包括以前流感传播和疫苗覆盖率的数据将使我们能够预测 即将到来的提前9个月以上的流感流行的强度和亚型/谱系分布。 这项拟议的研究将通过1)提高我们对相互作用的理解而有益于人类健康 人类免疫和病毒抗原漂移之间的关系以及2)提高流感的准确性和及时性 为流感预防和治疗分配资源留出更多时间。

SUMMARY/ABSTRACT This project will fill a fundamental knowledge gap in influenza epidemiology, which is the lack of a quantified relationship between viral antigenic drift and human susceptibility to influenza. We will then apply this new knowledge to improve the accuracy and timeliness of influenza forecasts. Antigenic drift refers to gradual changes in the surface proteins of influenza viruses, which allow new virus strains to escape acquired immunity and to re-infect individuals who were previously infected with influenza. Antigenic cartography can quantify the magnitude of antigenic drift (i.e. the antigenic distance) between influenza virus strains. To date, however, the relationship between antigenic distance and susceptibility to infection has not been quantified for human influenza. We will use a mechanistic model of influenza transmission and immunity to estimate the association between increasing antigenic distance and increasing susceptibility to infection with influenza. For this, we will take advantage of a unique data resource: active influenza surveillance conducted since the 2011/12 influenza season through the US Influenza Vaccine Effectiveness Network. These data include population-based estimates of the incidence of influenza, stratified by virus subtype/lineage and with antigenic and genetic characterization of circulating influenza viruses, in three geographically distinct US states. The data also include influenza vaccine coverage for the target populations. We will apply our mechanistic influenza model to these data and quantify the drift/susceptibility association. We will then apply these findings to improve forecasting of seasonal influenza epidemics. Two different approaches are currently taken to influenza forecasting. Short-term forecasts use near-real-time surveillance data to predict the timing and intensity of the peak in influenza cases, with lead times of a few weeks. Long-term forecasts use data on the relative prevalence of different influenza strains to predict which strains will dominate the upcoming season. At present neither short- nor long-term forecasting methods make effective use of data on pre-existing immunity to influenza due to vaccination or prior circulation of influenza strains. Having quantified the drift/susceptibility association, we will test the forecasting abilities of our influenza model. We hypothesize that including data on prior circulation of influenza and on vaccine coverage will allow us to forecast the intensity and subtype/lineage distribution of upcoming influenza epidemics with lead times of 9+ months. The proposed research will benefit human health by 1) improving our understanding of the interplay between human immunity and virus antigenic drift and 2) improving the accuracy and timeliness of influenza forecasts, allowing more time for the allocation of resources for influenza prevention and treatment.