Manore Carrie A, Davis Justin K, Christofferson Rebecca C, Wesson Dawn M, Hyman James M, Mores Christopher N
Center for Computational Science, Tulane University, New Orleans, Louisiana, USA.
School of Public Health and Tropical Medicine, Tulane University, New Orleans, Louisiana, USA.
PLoS Curr. 2014 May 30;6:ecurrents.outbreaks.f0b3978230599a56830ce30cb9ce0500. doi: 10.1371/currents.outbreaks.f0b3978230599a56830ce30cb9ce0500.
We have identified environmental and demographic variables, available in January, that predict the relative magnitude and spatial distribution of West Nile virus (WNV) for the following summer. The yearly magnitude and spatial distribution for WNV incidence in humans in the United States (US) have varied wildly in the past decade. Mosquito control measures are expensive and having better estimates of the expected relative size of a future WNV outbreak can help in planning for the mitigation efforts and costs. West Nile virus is spread primarily between mosquitoes and birds; humans are an incidental host. Previous efforts have demonstrated a strong correlation between environmental factors and the incidence of WNV. A predictive model for human cases must include both the environmental factors for the mosquito-bird epidemic and an anthropological model for the risk of humans being bitten by a mosquito. Using weather data and demographic data available in January for every county in the US, we use logistic regression analysis to predict the probability that the county will have at least one WNV case the following summer. We validate our approach and the spatial and temporal WNV incidence in the US from 2005 to 2013. The methodology was applied to forecast the 2014 WNV incidence in late January 2014. We find the most significant predictors for a county to have a case of WNV to be the mean minimum temperature in January, the deviation of this minimum temperature from the expected minimum temperature, the total population of the county, publicly available samples of local bird populations, and if the county had a case of WNV the previous year.
我们已经确定了1月份可用的环境和人口统计学变量，这些变量可预测次年夏季西尼罗河病毒（WNV）的相对规模和空间分布。在过去十年中，美国人类WNV发病率的年度规模和空间分布变化很大。蚊虫控制措施成本高昂，更好地估计未来WNV疫情的预期相对规模有助于规划缓解措施和成本。西尼罗河病毒主要在蚊子和鸟类之间传播；人类是偶然宿主。先前的研究表明环境因素与WNV发病率之间存在很强的相关性。人类病例的预测模型必须既包括蚊鸟疫情的环境因素，也包括人类被蚊子叮咬风险的人类学模型。利用美国每个县1月份可用的天气数据和人口数据，我们使用逻辑回归分析来预测该县次年夏季至少有一例WNV病例的概率。我们验证了我们的方法以及2005年至2013年美国WNV发病率的时空情况。该方法被应用于预测2014年1月下旬2014年的WNV发病率。我们发现，一个县出现WNV病例的最显著预测因素是1月份的平均最低温度、该最低温度与预期最低温度的偏差、该县的总人口、当地鸟类种群的公开可用样本，以及该县前一年是否有WNV病例。
