Department of Microbiology and Immunology, University of Michigan, Ann Arbor, MI, USA.
Bull Math Biol. 2020 Jun 13;82(6):78. doi: 10.1007/s11538-020-00752-9.
We present a framework for discrete network-based modeling of TB epidemiology in US counties using publicly available synthetic datasets. We explore the dynamics of this modeling framework by simulating the hypothetical spread of disease over 2 years resulting from a single active infection in Washtenaw County, MI. We find that for sufficiently large transmission rates that active transmission outweighs reactivation, disease prevalence is sensitive to the contact weight assigned to transmissions between casual contacts (that is, contacts that do not share a household, workplace, school, or group quarter). Workplace and casual contacts contribute most to active disease transmission, while household, school, and group quarter contacts contribute relatively little. Stochastic features of the model result in significant uncertainty in the predicted number of infections over time, leading to challenges in model calibration and interpretation of model-based predictions. Finally, predicted infections were more localized by household location than would be expected if they were randomly distributed. This modeling framework can be refined in later work to study specific county and multi-county TB epidemics in the USA.
我们提出了一个使用公共可用的合成数据集对美国县的结核病流行病学进行离散网络建模的框架。我们通过模拟密歇根州 Washtenaw 县的单一活动性感染在 2 年内的疾病传播，探索了这个建模框架的动态。我们发现，对于足够大的传播率，即活跃传播超过再激活，疾病流行率对在偶然接触（即不共享家庭、工作场所、学校或集体宿舍的接触）之间分配的传播权重很敏感。工作场所和偶然接触对活跃疾病传播的贡献最大，而家庭、学校和集体宿舍接触的贡献相对较小。模型的随机特征导致预测时间内感染数量的不确定性显著增加，从而对模型校准和基于模型预测的解释提出了挑战。最后，预测的感染在家庭位置上比随机分布时更集中。这个建模框架可以在以后的工作中进一步细化，以研究美国特定县和多县的结核病流行。
