基于时空传播动力学模型联合地理统计模型的传染病时空建模方法的开发与应用

在大数据背景下，构建数据驱动的传染病时空模型，是精准预测传染病动态时空分布、分析其重要影响因素及评价时空特异性防治措施的关键方法。传染病时空建模方法主要包括时空传播动力学模型和地理统计模型。前者不易进行高空间精度的数据拟合、难以进行过多影响因素的研究；后者对疾病的动态变化不敏感、长期预测能力有限、难以对不同干预策略的效果进行模拟。本研究拟针对当前时空建模方法的局限性，基于前期研究基础，以新型冠状病毒肺炎、手足口病、华支睾吸虫病和血吸虫病为典型，构建疾病传播、影响因素、干预策略及地点间相互作用的理论框架，研究时空传播动力学模型联合地理统计模型的构建机制。针对相应传染病的传播特征对模型进行细化，开发多层次联合模型，并分析不同类型传染病在模型构建中的共性与特性。将联合模型应用于相应传染病的大尺度、高时空分辨率的时空传播预测、影响因素分析及干预措施评价中，为疾病的防治提供新方法、新知识和新思路。

In the context of big data, data-driven spatial-temporal modeling is the state-of-art to accurately predict the dynamic spatial-temporal distribution of infectious diseases, to analyze their important influencing factors and to assess the spatial-temporal-specificity control measures. There are two major types of spatial-temporal modeling approaches for transmission of infectious diseases: the spatial transmission dynamic models and the geostatistical models. The former ones are difficult to fit the large amount of spatial data as well as to investigate a large number of potential influencing factors; the latter ones are insensitive to disease dynamics, thus limited for long-term forecasting, and is difficult to simulate the effects of different intervention strategies. The current proposed research aims to address the above limitations through development of hybrid modeling approaches combined with spatial transmission dynamic models and geostatistical models. Firstly, a theoretical framework will be developed, for disease transmission, influencing factors, intervention strategies, and inter-location interactions, based on which, the mechanisms for constructing the hybrid models will be studied. The hierarchy joint models will be developed by refining the characteristics of different infectious diseases into the primary modeling structure. Commonalities and characteristics of different types of infectious diseases will be investigated during the model development process. Based on prior research, four infectious diseases, that is coronavirus disease (COVID-19), hand, foot and mouth disease, clonorchiasis and schistosomiasis, are taken as applicable examples. The developed hybrid models will be used to predict the spatial-temporal dynamics, to study the influencing factors and to evaluate control strategies for the above diseases at high spatial-temporal resolution across large-scale, which will provide innovative methodology, up-to-date knowledge and new ideas for disease control and prevention.