台灣地區社會環境因子對女性乳癌空間分佈模式 之相關性分析(2005-2007年)

Abstract

目的：流行病學結合地理科學已經成為研究趨勢。本研究想透過疾病地圖與空間分析方法瞭解台灣地區女性乳癌在未滿40歲及40歲以上發生率、盛行率與死亡率之空間分佈情形並比較不同年份的時空變異，進而利用空間統計方法探索資料的空間特性，尋找乳癌群聚地區的社會環境致病因子。 方法：採橫斷式研究設計、次級資料分析，運用2005-2007年「癌症登記資料庫」、「死因檔資料庫」以及衛生署「鄉鎮癌症就診統計-乳癌」，研究對象為女性乳癌患者。空間分析以最近鄰居法(k-Nearest Neighbors)之最近五個鄉鎮(k=5)為鄰區的權重矩陣，病已全域型空間自相關指標Moran’s I與區域型空間自相關LISA進行空間相依性分析，探討乳癌是否呈現群聚現象。本研究並進一步探討社會環境因子與乳癌的相關性，利用空間延遲模型探討空間鄰近效應與利用地理加權迴歸解釋地理變異情形。 結果：台灣地區女性在2005-2007年三個年段之未滿40歲之乳癌準準化盛行率與40歲以上乳癌標準化發生率及盛行率有顯著群聚現象；2005-2007年死亡率在全年齡都沒有發現群聚現象。進一步針對群聚現象瞭解與社會環境因子之相關性，結果發現與都市化綜合發展、速食店密度和女性平均工時皆為顯著正向相關；而與工業區沒有統計顯著。利用空間延遲模型發現空間鄰近效應有達到顯著影響，因而使用地理加權迴歸解釋地理變異情形。在地理加權迴歸與傳統一般迴歸進行模式比較之後發現，地理加權迴歸之參數估計較傳統迴歸來的精確，顯示利用地理加權迴歸分析，可在控制地理變異下探討社會環境因子對乳癌的影響，可視為較佳分析模式。 結論：本研究利用空間分析瞭解台灣地區女性乳癌之社會環境危險因子，發現乳癌確實存在區位特性，希望相關單位針對乳癌防治能更因地制宜以達到有效預防的成效。

The combination of Epidemiology and Earth Sciences has become a research trend lately. The purpose of this research is to understand the occurrence rate, the prevalence of breast cancer and the death rate among Taiwanese women whom are below forty and above forty years old through disease mapping and spatial analysis regarding different years of spatial distribution and temporal and spatial variation. Thus, use spatial statistical method to explore the spatial characteristics of the data and to search for the social causative agent among breast cancer cluster region. The subjects were female breast cancer patient. Based on cross-sectional study design and secondary data analysis apply 2005~2007 Cancer Registry Database, Cause of Death File Database and Department of Health of Rural Area Cancer Treatment Statistics – Breast Cancer. Spatial analysis is based on K Nearest Neighbors method and five of the nearest towns (k=5) are adjacent areas’ spatial weight matrix. The disease has global spatial autocorrelation index, Moran’s I and regional spatial autocorrelation index, LISA, are as spatial dependency analysis to investigate whether breast cancer is cluster phenomenon or not. This research will further explore the correlation between the society environmental factor and breast cancer, the use of space delay model of spatial proximity effect and geographically weighted regression to explain the geographic mutation. Within 2005 to 2007, the three year period has shown that breast cancer under age of 40 quasi-standardized prevalence and over forty years old, breast cancer standardized incidence and prevalence of significant clustering phenomenon among Taiwanese women. However, there is no indication showing that death rate does not appear to be cluster phenomenon among all ages. To further understand clustering phenomenon of social environmental factor, the result shows that urbanization, fast food chain stores density, and average female working hours are significant and positive correlation. Though, there is no valid statistics of industrial area. It is found that the use of space delay model of spatial proximity effect has significant impact and because of that, geographically weighted regression has been used to explain the geographic mutation. After comparing geographically weighted regression and traditional regression model, geographically weighted regression has more accurate parameter estimation than traditional regression model. It is indicated that geographically weighted regression method can be seen as the best analysis model regarding investigate the impact of social environmental factor towards breast cancer while controlling geographic variation. This research uses spatial analysis to comprehend the relationship between breast cancer among Taiwanese women and social environmental risk factor. As a result, breast cancer does exist the characteristics of locational factor. Hope the relevant units for breast cancer prevention can be more adapted to local conditions in order to achieve effective prevention.