Study On Influencing Factors Of Incidence Of Bacillary Dysentery In Gansu Province Based On Spatial Regression Techniques

Objectives To explore the influencing factors of bacillary dysentery incidence in Gansu from 2009 to 2014.To reveal the relations between meteorological,socio-economic factors and bacillary dysentery incidence based on spatial regression models and bayesian space-time models.To provides a scientific basis for monitoring and prevention of bacillary dysentery in Gansu province.Methods1.The incidence of bacillary dysentery in 87 countries in Gansu was collected from “China Information System for Disease Control and Prevention”,including the bacillary dysentery cases and population sizes from 2009 to 2014.Meteorological data were collected from China meteorological data sharing network,including annual average temperature,annual average temperature difference,annual average relative humidity and annual rainfall.Meanwhile,Inverse Distance Weighted was applied to calculate meteorological index in other districts every year.Per-capita GDP and population density data were collected from statistical yearbook of Gansu Province,from 2010-2015.2.Ordinary least squares regression model,spatial lag model and spatial error model were applied to explore the relationship between the bacillary dysentery incidence and meteorological and socio-economic factors every year in Gansu.The most popular information criteria was used to select the optimal spatial regression model each year.3.In addition to the spatial autocorrelation of data,the spatial heterogeneity of bacillary dysentery was taken into account when examining the association of meteorological and socio-economic factors with bacillary dysentery incidence by local spatial regression model-geographically weighted regression model.4.Taking into account the time-space effect,bayesian space-time model was built to reveal the relation between the factors and bacillary dysentery incidence in Gansu.Results1.The results of the global regression model showed that the spatial lag model was better than the ordinary least squares regression model in 2009,and ordinary least squares regression model was found to be suitable for bacillary dysentery incidence in Gansu from 2010 to 2014 because there was no spatial autocorrelation.However,in terms of the evaluation index,the fitting effect of spatial regression model was slightly better than the ordinary least squares regression model.2.The results of the global regression model showed that:In 2009,per-capita GDP was positively correlated with the bacillary dysentery incidence in Gansu(P<0.001).In 2010,the bacillary dysentery incidence in Gansu was positively correlated with precipitation(P<0.001),and positively correlated with GDP per capita(P<0.001),and also positively correlated with the average temperature difference(P<0.05).In 2011,annual rainfall and GDP per capita were both positively related to the bacillary dysentery incidence of Gansu(P<0.005).In 2012,the bacillary dysentery incidence in Gansu was positively correlated with precipitation(P<0.001),and positively correlated with GDP per capita(P<0.005),and negatively correlated with the average relative humidity(P<0.05).In 2013,annual rainful was positively correlated with the bacillary dysentery incidence in Gansu(P<0.05).In 2014,the bacillary dysentery incidence in Gansu was positively correlated with precipitation(P<0.005),and positively correlated with GDP per capita(P<0.001).3.The results of geographically weighted regression model showed that the AICc value of geographically weighted regression model in 2009 was smaller than the ordinary least square regression model,which indicated that there exists spatial variability of bacillary dysentery incidence in Gansu in 2009.The results showed that there doesn't exist spatial variability of bacillary dysentery incidence in Gansu from 2010-2014.4.The results of bayesian space-time model showed that the DIC value of non-spatial and non-temporal model was far greater than spatiotemporal independent model,which indicated that the fitting effect of the spatiotemporal independent model was better than the non-spatial and non-temporal model.Meanwhile,the results showed that the bacillary dysentery incidence in Gansu was related to population density,annual rainfall and GDP per capita.Conclusions The result of this study suggested that there was a certain correlation between bacillary dysentery incidence of different regions of Gansu in 2009 and the spatial lag model was appropriate.However,ordinary least squares regression model was found to be suitable for bacillary dysentery incidence in Gansu from 2010 to 2014 because there was no spatial autocorrelation.Meanwhile,the result showed that there exists spatial variability of bacillary dysentery incidence in Gansu in 2009.The results of Bayesian space-time model indicated that it was better to reveal the relation between the factors and bacillary dysentery incidence in Gansu when time-space effect were taken into account.The results of this study showed that GDP per capita was related to bacillary dysentery incidence in Gansu in every year from 2009 to 2014(except 2013)and rainfall was related to bacillary dysentery incidence in Gansu from 2010 to 2014.Meanwhile,the results indicated that the bacillary dysentery incidence in Gansu in 2010 was related to the average temperature difference and the bacillary dysentery incidence in Gansu in 2012 was related to the average relative humidity.