用于潜在对数高斯结果的时空块聚合模型：在威尔士废水病毒浓度建模中的应用
A spatio-temporal block aggregation model for latent log Gaussian outcomes: application on modelling wastewater virus concentration in Wales
浏览论文内容
中文总结 AI 辅助
本研究针对废水监测数据的块聚合问题，提出潜在对数高斯结果的时空块聚合模型，采用INLA方法推断，应用于威尔士47个集水区的新冠病毒建模，证实其预测性能良好且与阳性率强相关，为公共卫生监测提供新方法。
中文摘要 AI 辅助
基于废水的流行病学已成为监测社区层面传染病动态的重要工具，能提供覆盖全人群的信号，补充临床监测。然而，废水测量值常为不规则空间单元上的聚合值。本研究开发了一种将潜在空间连续过程与聚合结果关联的方法，提出了针对潜在对数高斯结果的时空模型，为推断和预测提供连贯框架，允许对任意空间配置上的过程进行积分，该框架还可用于后续分析，如将废水信号与行政区域的健康结果关联。我们采用贝叶斯框架，通过线性化集成嵌套拉普拉斯近似（INLA）方法进行推断。将所提方法应用于建模2022年8月初至2023年7月底威尔士47个集水区废水中的SARS-CoV-2 N1基因拷贝数。结果表明，该模型能捕捉时空模式，具有良好的预测性能，且估计的病毒基因拷贝数与地方当局层面的COVID-19 PCR检测阳性率密切相关。研究结果强调，在分析废水监测数据时，明确建模块聚合的重要性，所提框架为将环境监测数据整合到公共卫生监测系统提供了灵活且有原则的方法。
英文摘要
Wastewater-based epidemiology has emerged as a valuable tool for monitoring community-level infectious disease dynamics, providing population-wide signals that complement clinical surveillance. However, wastewater measurements are often observed as aggregated values over irregular spatial units. This work develops an approach to link an underlying spatially continuous processes and an aggregated outcome. We propose a spatio-temporal model for latent log-Gaussian outcomes that provides a coherent framework for inference and prediction, allowing the process to be integrated over arbitrary spatial configurations. This framework can also be used for subsequent analyses, such as linking wastewater signal to health outcomes at administrative areas. We use a Bayesian framework for inference via the linearised integrated nested Laplace approximation (INLA) approach. We apply the proposed methodology to model SARS-CoV-2 N1 gene copies in wastewater across 47 catchment areas in Wales from the beginning of August 2022 to the end of July 2023. The results demonstrate that the model captures spatial and temporal patterns and has good predictive performance. Results also show that estimated viral gene copies are strongly linked to positivity rates from COVID-19 PCR tests at the local authority level. Our findings highlight the importance of explicitly modelling block aggregation when analysing wastewater surveillance data. The proposed framework provides a flexible and principled approach for integrating environmental surveillance data into public health monitoring systems.
