A Study Of Temporal-spatial Aggregation And Prediction Of Bacillary Dysentery

The bacillary dysentery is an Intestinal infectious disease caused by Shigella,with fever,diarrhea,and abdominal pain as the main symptoms.It is mainly transmitted through the feces-mouth^[1].It can be divided into acute and chronic dysentery according to the different course.Every year about 160 million people in the world are infected with Shigella,most of them from developing countries,and 1.1 million people died from bacterial dysentery^[2].Children under five account for two-thirds^[3].Bacterial dysentery is a Class B communicable diseases in China.Although the incidence has decreased from 1.3 million in 1991 to approximately 210,000 in 2012,due to the strong spread of the disease,the duration of immunization after infection is relatively short,and no effective vaccine in prevention,it is easy to infect again.therefore,we should continue to strengthen the prevention and treatment of bacterial dysentery^[4].Chongqing is dominated by hills and mountains.There are numerous rivers and abundant rainfall,which is conducive to the reproduction and spread of Shigella.In recent years,the incidence of bacterial dysentery is forefront of the statutory disease in Chongqing.The early warning of infectious diseases aims to find out the distribution of infectious diseases,predict trends in time,and establish prediction models so that predictions can be made before the outbreak of diseases to identify possible risks,and provide scientific basis for the early prevention and control measures^[5].Therefore,the effective early forecasting model of infectious diseases can help control the spread of the disease.In recent years,complete early warning systems have been established around the world^[6][7].Although there are many methods to establish early prediction models,most of them do not consider the possible influencing factors of infectious diseases.When the data are missing,many prediction models no longer applicable.At the same time,the occurrence and spread of infectious disease generally have temporal and spatial characteristics,Study of temporal and spatial clustering contributes to the development of disease prevention and control measures.there are some studies of infectious diseases with spatial analysis techniques^[19-21].The paper mainly uses spatial analysis technology to analyze the spatial and temporal aggregation characteristics of bacterial dysentery in Chongqing and the bayesian network model used to construct a risk predictive model of bacterial dysentery.The model not only considers the Influencing factors?meteorological factors,socioeconomic factors,and spatial factors?,but also applies when appear absence of data,Afterwards,the model can be improved based on the posterior knowledge,which is very suitable for early warning of infectious diseases.The study contain the following sections:First,describing the epidemiological characteristics of bacterial dysentery in Chongqing from2005 to 2016,and identify its epidemiological characteristics,using GIS to map the incidence rate to show its spatial distribution characteristics.Using Geoda software to study the spatial correlation of the incidence of bacterial dysentery and uses the spatio-temporal scanning statistics with SaTScan 9.4.2 software to analyze the temporal-spatial clustering of the bacterial dysenteric from 2005 to 2016 in Chongqing to explore its temporal-spatial distribution characteristics;finally selected 70%of the sample as a Training set,and established four Bayesian networks with R3.3.1 software to predict the risk of bacterial dysentery,and we use 30%of sample as a test set to compare the capability of the model,the Bayesian network is compared with Random forest and traditional Logistic regression model with Evaluation index such as the accuracy and AUC.The research results are following:?1?The average annual incidence of bacterial dysentery was 31.0437/100,000 with decline trend in Chongqing from 2005 to 2016.The incidence rate of males was higher than that of females.The incidence of bacterial dysentery was the highest among those under 5 years old?48,309 cases,accounting for 44.6%?.Occupation distribution,the scattered children occupy largest proportion?46097 cases,42.55%?with increasing trend.The incidence of bacterial dysentery had obvious seasonality,peaking from May to October?72103cases,accounting for 66.56%?.The area of high incidence of bacterial dysentery shows reduced trend in Chongqing,and is mainly concentrated in the the main urban districts,such as Shapingba District,Jiangbei District,Nan'an District and Jiulongpo District with the highest incidence rate?67.178/100,000?,and the lowest incidence rate?16.6765/100,000?in the northeastern districts.?2?The bacillary dysentery incidence showed a spatial autocorrelation during study period,High-High clusters distributed in main urban districts,Low-Low clusters and High-Low clusters presented in the one-hour economic circle and Northeast Area.?3?The temporal-spatial clustering analysis revealed that bacillary dysentery clusters most likely occurred in main urban districts from June to October,and Consistent with the spatial hotspots.?4?Using three methods to construct Bayesian network and compare with Random forest and traditional Logistic regression model,it is found that the Bayesian network model with comprehensive structure learning and domain knowledge has the best prediction performance.Structural learning found that the higher Aggregation level,the higher population density and per capita GDP,the higher average temperature,the greater precipitation,and the longer t sunshine hours,the risk level of bacterial dysentery will rise.In summary,the more economically developed metropolitan area is the key prevention and control area for the epidemic of bacterial dysentery in Chongqing.Because of high incidence of bacterial dysentery in summer and autumn,publicity efforts should be intensified to allow the public to have the correct hygiene habits and eating habits,while strengthening drinking water hygiene supervision.In addition,we should adopt targeted prevention and control measures for children,farmers,and migrant workers with health promotion to popularize knowledge on the prevention and treatment of bacterial dysentery.Meteorological factors,socio-economic factors are important factors are the risk factors of bacterial dysentery,and Bayesian networks with comprehensive structural learning and domain knowledge construction can better predict the risk of bacterial dysentery,it provides new ideas for the establishment of early infectious disease warning models.