疾病發生率地圖之穩定克里金法

A Stabilized Kriging Method for Mapping Disease Incidence Rates

許哲嘉(Che-Chia Hsu)

指導教授 ： 李文宗

國立臺灣大學/公共衛生學院/流行病學與預防醫學研究所/碩士(2020年)

https://doi.org/10.6342/NTU202002981

摘要

疾病發生率的地圖繪製是流行病學研究重要的一環，並能協助公共衛生的策略擬定。疾病發生率地圖多以縣市或鄉鎮市區為單位繪製。疾病發生率在人口數較少的地理區域較不穩定（因其有較大的變異數），容易出現極高或極低的數值。經驗貝氏法可以針對不同地理區域變異數的不同進行適當的處理。發生率因而得以穩定化。克里金法可以打破地理邊界的束縛，以等高線的形式呈現平滑曲面，但未有如經驗貝氏法之穩定效果。本研究提出穩定克里金法繪製疾病發生率地圖。穩定克里金法容許各縣市或鄉鎮市區有不同的誤差。蒙地卡羅模擬顯示穩定克里金法在所有情境下皆優於其他三種方法（原始發生率、經驗貝氏法及傳統克里金法）。台灣男性口腔癌發生率的實例研究發現，穩定克里金法地圖中，台灣中央山地地區的年齡標準化發生率已穩定化，不再有數值高低極大的落差。穩定克里金法地圖有較佳的解析度，能協助找出數個口腔癌發生率的熱點及冷點。我們推薦採用穩定克里金法進行疾病率地圖的繪製。

關鍵字

地圖 ； 克里金法 ； 平滑 ； 疾病發生率

並列摘要

The mapping of disease incidence rates is an important part of epidemiological research, which also help informs public health policy. Most of the disease incidence maps are drawn in units of counties, cities or towns. Disease incidence rates are not stable in regions of small population (because of large variance), and in those regions it is prone to appear extremely high or extremely low values. The empirical Bayes method can take different variances in different regions into proper account. Consequently, incidence rates can be stabilized by using this method. The Kriging method can break the constraints of geopolitical boundaries and produce a smooth curved surface in the form of contour lines, but it does not have the stabilizing effect as the empirical Bayes method. This study proposes a stabilized Kriging method to map disease incidence rates cancer. The stabilized Kriging method allows different errors in different counties/ cities. Monte Carlo simulations show that the stabilized Kriging method is superior to the other three methods (original incidence rates, empirical Bayes method, and traditional Kriging method) in all situations. Real-world data analysis about oral cancer incidence in men in Taiwan shows that the age-standardized incidence rates in the central mountainous region of Taiwan have been stabilized using our proposed method; there are no longer huge differences in numerical values. Additionally the stabilized Kriging map has better resolution and can help to find a number of hot spots and cold spots in the incidence of oral cancer. We recommend using the stabilized Kriging method for mapping disease rates.

並列關鍵字

mapping ； Kriging ； Smoothing ； Disease incidence rates

參考文獻

[1] Marshall RJ. Mapping disease and mortality rates using empirical bayes estimators. Journal of the Royal Statistical Society, Series C (Applied Statistics) 1991;40:283-294.

[2] Lin WC, Lin YP, Wang YC, Chang TK, Chiang LC. Assessing and mapping spatial associations among oral cancer mortality rates, concentrations of heavy metals in soil, and land use types based on multiple scale data. International Journal of Environmental Research and Public Health 2014;11:2148-2168.

[3] Diggle PJ, Heagerty P, Liang KY, Zeger SL. Analysis of longitudinal data, Oxford University Press; 2002.

[4] Beavis B, Dobbs I. Static optimization: optimization and stability theory for economic analysis, New York, NY: Cambridge University Press;1990:40.

[5] Chiles JP, Delfiner P. Geostatistics: modeling spatial uncertainty, New York, NY: Wiley interscience; 1999.