Poterek Marya L, Kraemer Moritz U G, Watts Alexander, Khan Kamran, Perkins T Alex
Department of Biological Sciences and Eck Institute for Global Health, University of Notre Dame, Notre Dame, IN 46556, USA.
Department of Zoology, University of Oxford, Oxford OX1 3SY, UK.
Pathogens. 2021 Feb 3;10(2):155. doi: 10.3390/pathogens10020155.
Measles incidence in the United States has grown dramatically, as vaccination rates are declining and transmission internationally is on the rise. Because imported cases are necessary drivers of outbreaks in non-endemic settings, predicting measles outbreaks in the US depends on predicting imported cases. To assess the predictability of imported measles cases, we performed a regression of imported measles cases in the US against an inflow variable that combines air travel data with international measles surveillance data. To understand the contribution of each data type to these predictions, we repeated the regression analysis with alternative versions of the inflow variable that replaced each data type with averaged values and with versions of the inflow variable that used modeled inputs. We assessed the performance of these regression models using correlation, coverage probability, and area under the curve statistics, including with resampling and cross-validation. Our regression model had good predictive ability with respect to the presence or absence of imported cases in a given state in a given year (area under the curve of the receiver operating characteristic curve (AUC) = 0.78) and the magnitude of imported cases (Pearson correlation = 0.84). By comparing alternative versions of the inflow variable averaging over different inputs, we found that both air travel data and international surveillance data contribute to the model's ability to predict numbers of imported cases and individually contribute to its ability to predict the presence or absence of imported cases. Predicted sources of imported measles cases varied considerably across years and US states, depending on which countries had high measles activity in a given year. Our results emphasize the importance of the relationship between global connectedness and the spread of measles. This study provides a framework for predicting and understanding imported case dynamics that could inform future studies and outbreak prevention efforts.
在美国，随着疫苗接种率下降以及国际间传播的增加，麻疹发病率急剧上升。由于输入性病例是在非流行地区引发疫情的必要因素，预测美国的麻疹疫情取决于预测输入性病例。为了评估输入性麻疹病例的可预测性，我们对美国的输入性麻疹病例与一个综合了航空旅行数据和国际麻疹监测数据的流入变量进行了回归分析。为了了解每种数据类型对这些预测的贡献，我们用流入变量的替代版本重复了回归分析，这些替代版本用平均值取代了每种数据类型，还用使用建模输入的流入变量版本进行了分析。我们使用相关性、覆盖概率和曲线下面积统计量（包括重采样和交叉验证）评估了这些回归模型的性能。我们的回归模型在预测给定年份特定州是否存在输入性病例方面具有良好的预测能力（受试者操作特征曲线的曲线下面积（AUC）=0.78）以及输入性病例的数量（皮尔逊相关性=0.84）。通过比较对不同输入进行平均的流入变量的替代版本，我们发现航空旅行数据和国际监测数据都有助于模型预测输入性病例数量的能力，并且各自有助于其预测输入性病例是否存在的能力。根据特定年份哪些国家麻疹活动高发，预测的输入性麻疹病例来源在不同年份和美国各州之间差异很大。我们的结果强调了全球连通性与麻疹传播之间关系的重要性。这项研究提供了一个预测和理解输入性病例动态的框架，可为未来的研究和疫情预防工作提供参考。
