Study On The Application Of GIS To Spatial Heterogeneity Of Disease

Objectives As traditional statistics that bases on the hypothesis of independence has defects in nature to process the spatial autocorrelation characteristics of disease distribution, the epidemiology study on disease distribution was limited. In this paper, theory and methods of spatial statistics are trialed on spatial analysis of primary hepatocellular carcinoma in Guangxi Zhuang autonomous region. The data of primary hepatocellular carcinoma has been collected from pathology diagnosed record, so as to explore the value of spatial statistics on studying spatial heterogeneity of disease, and look for scientific evidence for effective disease control and intervention.Methods The incidence data of primary hepatocellular carcinoma in Guangxi Zhuang antonomou region from 2000 to 2007 was collected. With geographic information system, several spatial statistics methods including exploratory spatial data analysis, semivariogram function, spatial autocorrelation, spatial scan statistic and Kriging interpolation were applied to spatial heterogeneity analysis of primary hepatocellular carcinoma. The trend of spatial distribution, hot spot and cold spot of incidence for primary hepatocellular carcinoma, relative risk of regions were detected and evaluated respectively. The quantitative analysis results for primary hepatocellular carcinoma in Guangxi has been visualized.Results 1.The fitting results of semivariogram function are as follow: nugget variance is 0.125, sill variance is 1.276, structural variance is 1.151, the ratio of nugget variance to sill variance is 0.098, the maximum range of spatial autocorrelation is 2.623(about 275 km), and goodness of fit is 0.914(R2=0.914). Such outcomes manifest that 90.2% of total variance is attributed to spatial autocorrelation.2.Kriging interpolation By distribution map, we find that the incidence of Liver Cancer in the southwest and south coastal of Guangxi is higher than in the north and northeast regions. The evaluation indices of prediction error are the following: Mean is 0.00016649, Root Mean Square is 0.2754, Average Standard Error is 0.2815, Mean Standardized is 0.001128, and Root Mean Square Standardized is 0.9822. The prediction error of cross validation indicates that the spatial distribution map of Liver Cancer in Guangxi is a good fitness.3. Analysis of spatial autocorrelation The global Moran's I coefficient is 0.5569 and has the statistical significance(P<0.001). The global Getis coefficient is 0.75 and also has the statistical significance(P<0.001). Both global spatial autocorrelation indicators illustrate that there is cluster with higher morbidity of Liver Cancer in Guangxi. According to LISA analysis, regions with higher morbidity cluster in the southwest of Guangxi(Z(Gi)>1.96), while regions with lower incidence locate in the north of Guangxi(Z(Gi)<-1.96).4.spatial scan statistic The outcome of purely spatial scan statistic is that the most likely cluster with higher morbidity is located in southwest of Guangxi(LLR=997.661, P=0.001, RR=4.918), including Fusui county,Long'an county,Wuming county,Pingguo county,Daxin county,Tiandeng county, Chongzuo city and Nanning city. The central coordinate of this cluster is 107.81520E,22.56770N, and the radius of such cluster is about 78.53 km. Conclusions Compared with traditional statistics analysis, Spatial Statistics is not only implement statistical inference to spatial heterogeneity on the hypothesis of spatial autocorrelation, but also has the distinctive advantage of locating cluster, quantifying the intensity of cluster, evaluating the relative risk of regions and identifying the pattern of spatial correlation. That information of disease provides scientific evidence for effective disease control and intervention; furthermore, and offers the prior knowledge for analysing the influential factors of disease cluster, by implementing spatial autocorrelation regression. In this paper, we reveal the distribution map of primary hepatocellular carcinoma in Guangxi, measure the radius of cluster with higher incidence, and find out the possible regions with lower incidence, by the spatial analysis of geographic information system. Such information provides the possible cue of etiological factor on the study of primary hepatocellular carcinoma in Guangxi Zhuang autonomous region.