台灣大腸癌死亡率之空間分析

Spatial analysis of colorectal cancer mortality in Taiwan

"目標：本研究旨在探究台灣大腸癌死亡率之空間分佈，檢驗空間鄰近效應及社會、醫療資源因子之相關性。 方法：使用2015年衛生福利部死因統計檔，對台灣349個鄉鎮市區（離島除外）大腸癌死亡率進行空間分析。以空間延遲模型探究大腸癌死亡率之空間鄰近效應。以地理加權廻歸分析鄉鎮市區大腸癌死亡率相關因子的異質性。 結果：大腸癌死亡率有空間聚集（Global Moran's I = .2182），熱區集中新北市、苗栗、嘉義縣及屏東。空間延遲模型不但呈現大腸癌死亡率之正向空間鄰近效應（廻歸係數 = .16）；亦具較佳的模型配適度（判定係數 = .3217）。空間延遲模型證實整體上，低教育程度比例、老年人口比例、大腸癌發生率愈高及家庭醫師人口比愈低則大腸癌死亡率亦愈高；但，地理加權廻歸估計顯示這些相關之鄉鎮市區估計具異質性，12.9%至79.1%鄉鎮顯著異於平均值。結論：地理資訊系統逐漸普及，未來地區資源分配，應考慮空間相關，評估在地風險及其異質性，以提升資源運用有效性及縮減地區健康不平等。（台灣衛誌2021；40(2)：225-240）"

Objectives: To explore the spatial patterns of colorectal cancer mortality in Taiwan, examine its neighborhood effect, and identify potentially associated factors. Methods: This study used spatial econometrics to analyze the geographical distribution of colorectal cancer mortality in Taiwan across 349 townships based on the 2015 cause of death statistics from Ministry of Health and Welfare. A spatial lag model was used to examine the neighborhood effects, and geographically weighted regression was applied to investigate the spatial heterogeneity in the relationship of social and medical resources with colorectal cancer mortality. Results: Colorectal cancer mortality exhibited significant spatial clusters (Global Moran's I = 0.2182), with hotspots in New Taipei city, Miaoli, Chiayi, and Pingtung. The spatial lag model yielded evidence of positive neighborhood effects ( = .16) with acceptable goodness of fit (coefficient of determination = .3217). Moreover, its results indicated that colorectal cancer mortality is associated with areas with higher proportion of limited education, higher proportion of older adults, higher incidence of colorectal cancer, and lower ratio of family physicians to the population. However, the geographically weighted regression revealed a heterogeneous association of these factors with mortality across 349 townships, with 12.9% to 79.1% of towns differing significantly from the mean estimates. Conclusions: With the increasing popularity of the geographical information system, policy makers focused on resource allocation should consider spatial correlation when identifying local risk factors and their heterogeneous effects on health or diseases. Doing so would improve resource effectiveness and reduce regional health inequalities. (Taiwan J Public Health. 2021; 40(2): 225-240)