机理知识和贝叶斯推理融合驱动的登革热传播推断与预测  

Inference and Prediction of Dengue Transmission Driven by Coupling Knowledge of Disease Transmission Mechanism and Bayesian Inference

作　　者：蔡晓琰 周佳仪 倪豪波 代婷婷 王聆溪 姚云冲 徐婷 肖丽娜 陈煜亮 郭貔[1] Cai Xiaoyan;Zhou Jiayi;Ni Haobo(Department of Preventive Medicine,School of Medicine,Shantou University,Shantou 515000)

机构地区：[1]汕头大学医学院预防医学系,515000 [2]南方医科大学南方医院

出　　处：《中国卫生统计》2024年第6期840-845,共6页Chinese Journal of Health Statistics

摘　　要：目的经典动力学机理模型基于传染病传播机理知识构建非线性微分方程组对其传播进行系统建模,其参数初始状态的估计误差会伴随着系统的迭代和演化而扩大,具有参数初值敏感依赖的局限,从而削弱模型预测能力。方法本研究采用贝叶斯推理方法与疾病传播动力学机理模型进行融合设计,通过数据同化技术实现模型参数的不断迭代更新和优化,克服经典动力学模型的参数初值敏感依赖的局限。基于此,本研究构建机理知识和贝叶斯推理融合驱动的登革热传播推断与预测框架SIR-EAKF,并将该融合驱动模型应用于广州市登革热传播的推断与预测。结果本研究构建的SIR-EAKF框架优化了集合模拟的状态参数,实现了对疾病传播参数的准确估计,从而使得动力学机理模型的集合预报更精确,能够提前准确地预测登革热在人群中传播和演化的趋势。结论基于以上融合模型可以实现对登革热流行和暴发的近似实时预测和追踪,提高人们对传染病疫情的早期应对和感知能力,为公共卫生防控争取更多宝贵的时间。Objective The classical dynamical mechanism model constructs a system of nonlinear differential equations based on the knowledge of the transmission mechanism of infectious diseases to systematically model their propagation,and the estimation error of the initial state of its parameters will expand along with the iteration and evolution of the system,which has the limitation of sensitive dependence on the initial value of the parameters,and thus weakening the predictive ability of the model.Methods In this study,a Bayesian inference method is used to integrate with the disease transmission dynamics model,and the data assimilation technique is used to achieve the continuous iterative updating and optimization of the model parameters,which overcomes the limitation of sensitive dependence on the initial value of parameters in the classical dynamics model.In this study,we constructed a fusion-driven dengue fever transmission inference and prediction framework,SIR-EAKF,which is driven by the fusion of mechanistic knowledge and Bayesian inference.This study introduces the basic principles,construction steps,analysis methods,and related considerations of the method,and applies the fusion-driven model to the inference and prediction of the spread of dengue fever in Guangzhou.Results The results show that the SIR-EAKF framework constructed in this study optimizes the state parameters of the ensemble simulation,and it is able to achieve an accurate estimation of the disease spread parameters,which makes the ensemble forecast of the dynamicalmechanism model more accurate and predicts accurately in advance the trend of dengue fever spreading and evolving in populations.Conclusions Based on the above fusion model can achieve approximate real-time prediction and tracking of dengue fever epidemics and outbreaks,which improves people's ability to respond to and sense infectious disease outbreaks at an early stage,buying more valuable time for public health prevention and control.

关 键 词：动力学机理模型 贝叶斯推理 卡尔曼滤波 登革热 

分 类 号：R195.1[医药卫生—卫生统计学]