趨勢面分析法在癌症地圖繪製上的應用：以臺灣的乳癌死亡率為例

Trend Surface Analysis for Cancer Mapping: Application to Breast Cancer Mortality in Taiwan

廖勇柏(Yung-Po Liaw) ； 李文宗(Wen-Chung Lee) ； 陳建仁(Chien-Jen Chen)

《中華公共衛生雜誌》 17卷6期 (1998/12) Pp. 474-484

https://doi.org/10.6288/CJPH1998-17-06-05

摘要

疾病的地理變異，是流行病學的重要課題。然而，傳統地圖繪製方法，以行政區域為單位常無法清楚看出地理分佈的趨勢，且常面臨死亡率不穩定，干擾地圖判讀的問題。本文利用一種統計技術稱為趨勢面分析，及結合地理資訊系統進行癌症地圖繪製。本研究以1982~1991年臺灣本島346個行政區域之乳癌死亡資料為例進行示範，並與傳統地圖繪製方法比較。結果顯示，透過一次趨勢面分析、二次趨勢面分析至五次趨勢面分析之動態變化，可一目瞭然看出死亡率聚集擴散之傾向，以往地廣人稀鄉鎮死亡率劇烈變動之現象，亦不復可見。相對危險性較高之地區清楚地集中在大都會區，包括台北縣、市，台中縣、市，台南市以及高雄市。反之，西南沿海及山地鄉，其相對危險性則較低。本文作者期待趨勢面分析法在公共衛生領域之應用能更加開展普及。

關鍵字

趨勢面分析 ； 地理資訊系統 ； 卜瓦松迴歸 ； 乳癌

並列摘要

The geographic variation in the occurrence of diseases is an important issue in epidemiology. The traditional mortality mapping technique, which is based on administrative areas, suffers from the problem of instability and has difficulty in identifying spatial patterns and trends of potential interest. In this paper, the authors apply the trend surface analytic technique to model mortality rates and then use the ”Geographical Information System” (GIS) to draw maps. Data of female breast cancer mortality in the 346 administrative areas in Taiwan from 1982 to 1991 were used as an example. As compared to the traditional method, it is found that the trend surface analytic technique stabilizes mortality rates especially in areas with small populations, such as in the mountainous areas. Moreover, the technique, with the building of the 1st to the 5th order models, can provide a dynamic exhibition of patterns and trends which leads to clues about clustering and diffusion process in the mortality. The study finds higher breast cancer mortality to be clustered in several main metropolitan areas including Taipei City/County, Taichung City/County, Tainan City and Kaohsiung City. The southwestern coastal areas and the mountainous areas reveal lower mortality. It is hoped that the trend surface analysis becomes a standard public-health tool in the days to come.

並列關鍵字

trend surface analysis ； geographical information system ； Poisson regression ； breast cancer