基于地理加权回归模型的我国女性肺癌发病空间影响因素分析

Analysis of spatial factors of women's lung cancer incidence in China: a geographically weighted regression model approach

目的应用地理加权回归模型探讨气象因素和大气污染因素影响我国女性肺癌发病的空间属性。方法从《2012中国肿瘤登记年报》获取2009年我国72个肿瘤登记点的肺癌发病数据,分别从中国气象局和《2008中国环境年鉴》获取2008年气象和大气污染资料。使用因子分析将收集的气象、大气污染等空间因素降维,消除多重共线性。以标准化发病率作为因变量、城乡类型及空间因素因子得分作为自变量,分别拟合泊松回归模型和地理加权回归模型。结果东北地区尤其是辽东半岛为我国女性肺癌的高发区域,具有明显的地区集聚性。城乡类型、纬度指向因子(PM10、温度、降水量)与女性肺癌发病的关联具有统计学意义,农村相对于城市为保护因素(RR=0.758,P=0.003);纬度指向因子为危险因素,因子得分越高(PM10越高,温度越低,降水量越小),女性患肺癌风险越大(RR=1.104,P=0.027)。地理加权回归模型结果进一步显示,城乡类型对我国女性肺癌发病的影响强度存在东西地区差异,纬度指向因子影响强度存在南北地区差异。结论地理加权回归模型对具有空间自相关性的数据获得更优的拟合,可以较好地揭示空间因素在地区间作用的差异。

Objective To explore the spatial characteristics of meteorological factors and air pollutants＇ impact on women＇s lung cancer incidence in China based on a geographically weighted regression model. Methods Factor analysis was used to perform dimension reduction, eliminate muhicollinearity of the collected spatial data. A Poisson regression model and a geographically weighted regression model were fitted in which age-standardized incidence was the independent variable and meteorological factors and air pollutants＇ levels are dependent variables. Results In northeast China, especially Liaodong Peninsula, high women＇s lung cancer incidence was discovered, which revealed regional clustering in such areas. Urban-rural attributes and latitude-related factor （PM10, temperature and rainfall） were found statistical significantly related to women＇s lung cancer incidence; Rural area was a protective factor in contrast to urban area （RR=0.758, P=0.003）, and the increase of the latitude-related factor score was a hazard factor （RR=1.104, P=0.027）. The results of geographically weighted regression model showed west-east differences in urban-rural attributes＇ impact and north-south differences in latitude-related factor＇s impact. Conclusion Geographically weighted regression can achieve better fitting for modelling data with spatial autocorrelation and better explanation of the difference in spatial factors＇ impact on different locations.