時間序列自我迴歸模型應用於1998年-2006年臺南市與高雄市登革熱案例之研究

A Research on the Analysis of Dengue Fever in Tainan City and Kaohsiung, Taiwan during the Epidemics of 1998-2006 using Autoregressive Models

江麗玲(Li Ling Jiang)

指導教授 ： 李正宇博士

亞洲大學/資訊學院/生物資訊學系碩士在職專班/碩士(2008年)

摘要

本研究利用時間序列分析登革熱的實際病例，找出登革熱的傳播模型、時間上的特性及限制因素；再利用交叉相關分析，計算登革熱病媒蚊指數與病例的關係及各區相關性。2002年高雄市登革熱病例暴增，因此本研究以1998年至2001年及2003年至2006年臺南市和高雄市的登革熱病例進行時間分段分析，利用時間序列自我迴歸（AR）模型分析2002年登革熱大流行前後傳染病行為的變化；本研究發現2002年以後登革熱傳播的模型正向強度有增加趨勢，而模型負向衰退的強度並沒有明顯改變，顯示登革熱在臺南市和高雄市的傳播情況越來越嚴重，然而傳染病的控管並未有明顯進步。此外，依照南區與高屏區登革熱病例數計算區域相關性，結果呈現不相關。而相鄰的臺南市與臺南縣，高雄市與高雄縣呈現正相關，顯示登革熱傳播應具有較強的區域性。在病媒蚊指數與登革熱病例相關性研究中，臺南市存在正相關。當臺南市病媒蚊指數增加，登革熱的可能病例數於四至五個月內跟著增加；此結果與新加坡研究有異。Joseph R Egger等研究1966-2002新加坡的登革熱資料，用2個數學模型研究傳播的強度、發生率和病媒蚊的繁殖。發現病媒蚊指數明顯下降，但是登革熱病例卻不斷的增加；這是未來值得探討的研究主題。

關鍵字

病媒蚊指數 ； 時間序列分析 ； 自我迴歸 ； 相關性 ； 登革熱

並列摘要

In this study, dengue fever epidemics in southern Taiwan occurred in 1998-2006 were analyzed using autoregressive models to find their representative epidemic models, temporal characteristics, and constraints. Furthermore, to find the correlation between the Aedes household index and the number of cases among administrative districts, cross correlations were evaluated and compared. Because of abrupt increase in the number of cases in 2002, to characterize the difference between before- and after 2002, the dataset was segmented into two and analyzed separately, i.e. 1998-2001 and 2003-2006, at Kaohsiung city and at Tainan city.
Behaviors of the epidemics before and after 2002 were analyzed and compared using autoregressive models (AR) to find the change before- and after year 2002. We found that the strengths of the epidemic spreading become stronger after 2002; in contrast, there is no corresponding control strategy on the epidemic elimination. It means that the epidemic is deteriorated, however, the efficacy of the epidemic control did not improve much after 2002 in both Kaohsiung city and Tainan city.
On the other hand, the cross correlation analyses showed that there is no significant correlation between southern area and Kaohsiung-Pingtong area. However, there exist highly positive correlations between Tainan city and Tainan county, and Kaohsiung city and Kaohsiung county. It suggests that the dengue fever epidemics in southern Taiwan are highly localized. As to the correlation between the Aedes household index and the number of cases, our results showed there exists positive correlation at Taiwan city. The cross correlation analysis showed that whenever the Aedes household index increases, the number of cases climbing for next consecutive four or five months, which contradicts with a similar study in Singapore. This shall be an interesting topic for further study.

並列關鍵字

Dengue Fever ； Aedes Household Index ； Time Series ； Correlation

參考文獻

行政院衛生署疾病管制局，2004 年12 月，傳染病防治工作手冊，台北：行政院衛生署疾病管制局出版。

行政院衛生署疾病管制局，2006 年7 月，登革熱防治工作手冊，台北：行政院衛生署疾病管制局出版

Box.G.E.P. and Jenkins,G.M.（1976）.Time Series Analysis Forecasting and Control Revied ,San Francisco

Cheng-Yu Lee.2005. Space-Time Modeling and Application to Emerging Infectious Diseases ,Michigan State University , East Lansing

H. Akaike. (1973). Information theory and an extension of the maximum likelihood principle. 2nd International Symposium on Information Theory, 267-281.