Hu Yi, Xiong Cheng-Long, Zhang Zhi-Jie, Bergquist Robert, Wang Zeng-Liang, Gao Jie, Li Rui, Tao Bo, Jiang Qiu-Lin, Jiang Qingwu
Geospat Health. 2013 Nov;8(1):125-32. doi: 10.4081/gh.2013.60.
When modelling prevalence data, epidemiological studies usually employ either Gaussian, binomial or Poisson models. However, reasons are seldom given in the literature why the chosen model was felt to be the most appropriate. In this study, we compared all three models for fitting schistosomiasis risk in Xingzi county, Jiangxi province, People's Republic of China. Parasitological data from conventional surveys were available for 36,208 individuals aged between 6 and 65 years from 42 sampled villages and used in combination with environmental data to map the spatial patterns of schistosomiasis risk. The results show that the Poisson model fitted the data best and this model identified the role of environmental risk factors in explaining the geographical variation of schistosomiasis risk. These factors were further used to develop a predictive map, which has important implications for the control and eventual elimination of schistosomiasis in the People's Republic of China.
在对患病率数据进行建模时，流行病学研究通常采用高斯模型、二项式模型或泊松模型。然而，文献中很少说明选择某种模型被认为是最合适的原因。在本研究中，我们比较了这三种模型对中华人民共和国江西省星子县血吸虫病风险的拟合情况。来自常规调查的寄生虫学数据可用于42个抽样村庄的36208名6至65岁的个体，并与环境数据相结合以绘制血吸虫病风险的空间模式图。结果表明，泊松模型对数据的拟合效果最佳，该模型确定了环境风险因素在解释血吸虫病风险地理变异方面的作用。这些因素进一步用于绘制预测图，这对中华人民共和国血吸虫病的控制和最终消除具有重要意义。
