估計局部空間自相關呈現傳染病擴散特徵的空間結構：比較2015與2023年台南登革熱疫情擴散

Abstract

疾病傳播涉及複雜的空間擴散模式，傳統空間分析研究經常使用鄰近矩陣或是人流資料代表疾病擴散的空間結構，但這些事先給定的結構可能無法捕捉到真正的傳播關係。本研究採用條件自適應自迴歸模型( Conditional Autoregressive Adaptive Model )估計傳染病傳播的空間擴散模式，並使用2015年和2023年台南市登革熱疫情的案例進行分析。研究結果說明，模擬資料的分析結果確定了模型對傳播關係的準確估計能力，而套用於登革熱實際資料的模型分析結果顯示：模型估計出的地區關係以及其所反映的擴散關係與過往文獻描述的擴散趨勢吻合，且另外找到了過去文獻未曾觀察到的潛在遠距離擴散關係。此外，將估計出的空間結構轉換為網絡並進行分群後，可以發現兩年度的分群結果相似，且地區之間的互動關係可能與傳統分區有關。這說明聚落之間的互動關係具有穩定性和持續性，有助於在疫情初期快速找到高風險區域並實施相對應的措施。這項研究的成果能為疾病傳播的研究提供新的想法，透過模型估計空間權重矩陣，可以同時捕捉擴張型擴散和遷移型擴散的結構特徵，打破過去研究必須事先給定空間結構的限制，並增強我們對疾病空間擴散特性的理解。若能確實掌握將地區之間的擴散以及互動的空間結構，對於疾病預防和控制策略的將具有一定的參考價值。

Disease transmission involves complex spatial diffusion patterns. Conventional spatial analysis often relies on spatial adjacency matrices or human mobility matrices, which may fail to capture true transmission relationships. This study employs a conditional autoregressive adaptive model to estimate the 2015 and 2023 dengue fever outbreak in Tainan, Taiwan, aiming to provide insights into spatial structure of disease diffusion. Through the model's estimation, we can simultaneously capture the distinctive structural characteristics of various diffusions, breaking free from the constraints of having to predefine spatial structures as required in previous studies. Simulated data confirms the model can accurately estimate transmission relationships. Analysis of actual data indicates the dengue outbreak exhibited both expansion diffusion to nearby districts and potential relocation diffusion to more distant regions. Furthermore, converting the estimated spatial structures into networks and performing clustering also helps to understand the interactions between regions. The results reveal that the clustering outcomes are similar for both years, and the interactions between regions may be related to traditional administrative divisions. The study's results offer new insights into the spatial diffusion characteristics of diseases by estimating spatial weight matrices through the model. This approach overcomes the limitation of having to predefine spatial structures and enhances our understanding of disease transmission dynamics. Understanding the estimated spatial structure and interactions between regions can enhance our comprehension of the spatial diffusion characteristics of diseases and provide critical references for formulating effective epidemic prevention and control strategies.