Modelling The Malaria Transmission And Risk Prediction Of Malaria In Yunnan Border Area

ObjectiveMalaria is one of the important parasitic diseases that seriously threatened human health safety,which also affect social and economic development.It has been become a global public health problem.In 2010,China Malaria Elimination Action Plan has been launched in the whole country,it means that the malaria control work has entered the elimination phase.The Yunnan border area is not only a malaria surveillance area,but also a main area for implementing malaria elimination program.To promote the implementation of Malaria Elimination Action Plan and improve the surveillance system about malaria.In this paper,we explore the malaria epidemic in the border areas of Yunnan,attempt to apply the time series analysis methods in malaria,analyze the spatial-temporal distribution patterns of Plasmodium vivax malaria and find dynamics trends over time.Then we establish a formula to detect the relations between meteorological factors and malaria.In Yunnan Border Area,we present a Bayesian geographic statistical modelling framework that can be used to estimate the effects of imported malaria on relative risk in different regions.Here,our main purpose is to provide some scientific basis evidence for the deployments of medical resources and the elimination of malaria in Yunnan border areas.Method1.Using the multi-time series state-space model to analyze the spatial patterns about local malaria P.vivax in Baoshan City from 2007 to 2010.The optimal model is selected according to the Akaike information criterion.The Shapiro method is used to test the white noise of the model residuals.2.The Bayesian geographical statistical model is constructed by imported cases to calculate the relative risk of local malaria in Baoshan region.The model parameters is simulated by Markov Chain Monte Carlo algorithm.3.To reveal the temporal patterns of P.vivax from 2006 to 2012 in Yunnan Tengchong,the seasonal and dynamic trend are performed on the half monthly time series analysis method.4.The dynamic linear model are employed to model temporal patterns about P.vivax in Tengchong city.The parameters in model are estimated by Kalman filter algorithm,the plots of fitted and predicted value are calculate by best fit model.5.The notion of vecorial capacity and meteorological factors are used to estimate the transmission of malaria and local P.vivax,both using linear and non-linear model.According Akaike information criterion to select best fitted models,then draw the predict malaria values of different models together.6.The linear regression model is used to test the structural changes in time series malaria data in Tengchong city from 2006 to 2012.By the minimum Bayesian information criterion,choosing the break points in time series data,the model residual is run in Augmented Dickey Fuller test.Result1.From 2007 to 2010,three different spatial patterns about P.vivax were find in in Baoshan region.Among them,Tengchong and Shidian city were independent with each other,the malaria situation in Changning,Longyang,and Longling counties were the sampling from one malaria populations.The rates of malaria in three cluster were all negative.2.In 2007-2010,the local P.vivax malaria in Baoshan region was related to imported cases.In the Bayesian geographical statistical model,the coefficients of imported cases became bigger in 2009.The malaria relative risk in Tengchong is the highest place in Baoshan.3.In 2006 to 2012,the P.vivax cases in Tengchong,showed a variation from year to year.The trends about malaria was negative.The period of malaria was existed every year,the cumulative of cases reached to their peaks in May to June.4.The dynamic linear model about 2006-2012 malaria time series data showed,the mean square error was 0.417.After excluded the effect of the time series data and the unit of measurement,the average absolute percentage error was 0.152.By Shapiro test performed on the residuals of the model(W=0.98,p=0.106),the unit root hypothesis could not rejected.5.The risk factors of malaria in Tengchong included vecorial capacity and average relative humidity.A non-linear model combining vecorial capacity and average relative humidity could explain 32.8% of malaria cases in this region.For non-linear model in local malaria data,the explanation could reach to 45.8%.6.The linear regression model in 2006-2012 about malaria data in Tengchong city had the lowest Bayesian information criterion,when the break point located in 2.Suggested that malaria decrease deviations consisted by three phases.Before August 2006,P.vivax was at a high level.In the following time to September 2009,the epidemic was decreased to second level.From September 2009 to December 2011,the epidemic was down to a low level.ConclusionFor our study,a novel method of multi-time series state-space model was used to identify the spatial patterns of malaria and decline trends in Baoshan.It was found that Tengchong city was a high-risk area for malaria;when focused on P.vivax malaria in Tengchong.There existed a clear period of malaria epidemic.The trend of disease in dynamic linear model is consistent with real data in time.So for short-term data,especially those combined with periodic characteristics,could achieve better prediction results in this model.The generalized additive model could be use to describe the nonlinear relationship between malaria epidemic and meteorological factors.From nonlinear model we found vectorial capacity and average relative humidity were the risk factors for malaria in Tengchong city.The time series regression model could test for structural change of malaria in times and presented how these change have been realized in real world,we could tell malaria fluctuations for specific time in mathematical.