檢定在傳染病熱區邊緣的風險增長趨勢：以高雄登革熱爆發為例

控制傳染性疾病的傳播與在適合的地區進行干預一直是近年來重要的議題，在過去的研究中，將疾病的熱區作為目標干預地，並針對熱區進一步進行後續的地理相關性研究是目前常用的方法。然而只針對熱區進行治理的政策往往會忽略了熱區周邊具有與熱區鄰近性高且混雜著比熱區更高比例可感宿主的區域「邊緣區」，這樣的特性可能會使得邊緣區成為更加危險甚至帶動疾病的傳播。因此，本研究的目的為在疾病爆發期間，畫定疫情的熱區與邊緣區，並探討邊緣區疾病風險的增加率是否會高於熱區本身的增加率。為了回答這個問題，我以長期為登革熱疫區的熱帶城市──台灣的高雄為例，使用1998年至2020年的資料。我使用條件自回歸模型（conditional autoregressive model, CAR模型）以研究區中每個里在不同時期的數據，對嚴重年、平常年的熱區、顯著邊緣區與非顯著邊緣區進行分析，兩種邊緣區接相鄰於熱區，但不顯著邊緣區的危險程度較顯著邊緣區來的高。接著，使用雙重差分（difference-in-difference, DID）回歸模型來比較兩兩區域間風險增長的趨勢。結果顯示，無論是嚴重年或平常年顯著邊緣區危險程度的改變率皆比熱區有更高的上升幅度；而只有在嚴重年有出現不顯著邊緣區危險程度的上升幅度比熱區來的高。另外，兩種邊緣區之間則不具有顯著的差異存在。整體而言，在熱區與兩種邊緣區的比較中，DID的迴歸係數為正值，代表邊緣區確實有比熱區具有更大的風險增長趨勢。本研究在實務上建議政策制定者應更加關注傳染病熱區的邊緣，並同時在學術上為疾病管制提供新的想法。

關鍵字

傳染病熱區；傳染病邊界；熱區邊緣；條件自回歸模型；雙重差分；登革熱

並列摘要

Controlling the spread of infectious diseases and intervening in the right regions have long been important issues in recent years. In past studies, identifying disease hotspots as the targeting places and conducting follow-up geographical correlation studies of the hotspots are the most used ways. However, the edge area, which is the periphery around a hotspot, is often ignored but has the characteristics where more susceptible individuals are located. Proximity to hotspots and a higher proportion of susceptible hosts may let the edge area become risky and even spread the disease. Thus, the purpose of this study is to delineate the hotspot and its edge area and understand whether the growth rate of disease risk would be higher in the edge area than in the hotspot during an epidemic. To answer this question, dengue in Kaohsiung, Taiwan, from 1998 to 2020 is taken as a case study because Kaohsiung is representative of a tropical city for dengue. The conditional autoregressive model (CAR model) is used to analyze period data for each village to derive the hotspot area, significant edge area, and non-significant edge area, respectively, in the selected serious years and normal years. And, then, using difference-in-difference (DID) regression to determine the growth of risk between each of the two areas. The results show that there is significant growth of risk in significant edge areas, more than in hotspot areas among all selected years, whereas the greater growth of risk in non-significant edge areas than in hotspot areas is only found in normal years, and a significant difference does not exist between two edge areas. Overall, the coefficients of DID regression are positive in comparison between hotspot areas and two types of edge areas, indicating that edges have larger trends of risk growth than hotspots. This study suggests policymakers should pay more attention to the edges of infectious disease hotspots and provides an insightful view of disease control.

並列關鍵字

infectious disease hotspot ； infectious disease boundary ； edges of hotspot ； conditional autoregressive model ； difference-in-difference ； dengue

參考文獻

Abadie, A. (2005). Semiparametric Difference-in-Differences Estimators. The Review of Economic Studies, 72(1), 1-19. https://doi.org/10.1111/0034-6527.00321

Google Scholar

Anderson, C., Lee, D., Dean, N. (2014). Identifying clusters in Bayesian disease mapping. Biostatistics, 15(3), 457-469.

Google Scholar

Anno, S., Hara, T., Kai, H., Lee, M.-A., Chang, Y., Oyoshi, K., Mizukami, Y., Tadono, T. (2019). Spatiotemporal dengue fever hotspots associated with climatic factors in taiwan including outbreak predictions based on machine-learning. Geospatial health, 14(2).

Google Scholar

Anselin, L. (1995). Local indicators of spatial association—LISA. Geographical Analysis, 27(2), 93-115.

Google Scholar

Antonio, F. J., Itami, A. S., de Picoli, S., Teixeira, J. J. V., Mendes, R. d. S. (2017). Spatial patterns of dengue cases in Brazil. PLOS ONE, 12(7), e0180715.

Google Scholar

主題瀏覽