Research On Spatio-temporal Distribution And Prediction For Malaria Epidemics In Hubei Province, China

Objectives:This study analyzed the spatial-temporal distribution of local malaria in Hubei province, to identify the high-risk areas and periods of malaria prevalence. The effect of meteorological factors on local malaria prevalence in Hubei province was evaluated. The meteorological forecasting model of malaria incidence was established. Using the ARIMA(Autoregressive Integrated Moving Average) model to predict the monthly incidence trend of local malaria in Hubei province. This research provided decision support for monitoring and prediction of local malaria in Hubei province finally provided scientific reference and theory guidance for guiding the prevention control of malaria and rational allocation of health resources.Methods:(1) The long-term development and change trend of local malaria epidemic situation in Hubei Province during2004-2011using the Cochran-Armitage trend test. The annual local malaria incidence distribution map was drawn in Hubei province during2004-2011.(2) Whether the spatial autocorrelation existed in the whole study area using the global Moran’s I spatial autocorrelation analysis of ArcGIS10.1software. The high-risk areas of local malaria prevalence was determined in Hubei Province during2004～2011using two local spatial cluster research methods including ArcGIS10.1local Moran’s I spatial autocorrelation analysis method and purely spatial scanning statistics method. The time and space-time distribution characteristics of malaria were studied so as to determine the high-risk periods and areas using purely temporal scanning, spatio-temporal scanning.(3) The correlation between different sizes of malaria incidence and meteorological factors was analyzed using Spearman rank correlation analysis. The meteorological factors influencing the incidence of malaria were selected using multiple regression stepwise regression analysis method. The regression fitting of meteorological factors was performed on the incidence of malaria.(4) The malaria prevalence in Zaoyang County was divided into low, medium and high-incidence according to constituent ratio of monthly number of patients accounting for annual total number of patients in this study. The future malaria prevalence in Zaoyang County was discriminated using meteorological factors by stepwise discriminant analysis method.(5) ARIMA model was constructed using local malaria incidence in Hubei province during2004～2009. The data were used for model test from January to December2010. The model fitting and prediction effects were evaluated.Results:(1) The incidence of local malaria showed an overall significant downward trend in Hubei province during2004～2011. Global spatial autocorrelation analysis results indicated that there was a certain spatial cluster in local malaria prevalence in Hubei province. A total of11high-risk counties were determined through Local Moran’ s I analysis from2004to2011, the median annual malaria incidence of high-risk counties decreasing was from58.81/100,000individuals in2004to0.79/100,000individuals in2011.The method of purely spatial scan statistics identified different11significant spatial clusters between2004and2011(eight most likely clusters and three secondary likely clusters). The space-time clustering analysis determined that the most likely cluster included13counties, and the time frame was from April2004to November2007.(2) The periodical changes of malaria incidence was obviously correlated with climate periodical changes in two regional scales between Hubei province and Zaoyang county during2004-2009. The correlations with temperature-related indicators and rainfall were more significant. The correlation coefficient was about0.7.78.1%of the change of monthly malaria incidence in Hubei Province was attributed to the average temperature of current month, last month (MeanT-o1) and average minimum temperatures of last2months (MinT.2);67.8%variation of monthly malaria incidence may attribute to average maximum temperature of current month and last month (MaxT-01) in Zaoyang County.(3) The prevalence of malaria in Zaoyang County was distinguished by establishing discriminant function using the meteorological factors in this study. Firstly32meteorological factors were distinguished using stepwise discriminant analysis. MinT0, MaxT0and D-012were introduced in the discriminant equations finally.73.61%of original grouped months were correctly classified by the discriminant function based on these three variables.(4) The model was established using the monthly local malaria incidence in Hubei Province during2004～2009. The results showed that the fitting effect of ARIMA (1,1,1)(1,1,0)12model was relatively optimal. The predicted incidence trend was completely consistent with the actual incidence trend, the actual values were in95%confidence interval of the prediction value. This showed that the model prediction values were consistent with the actual situation and the fitting effect was good.Conclusions:(1)The incidence of local malaria showed an overall significant downward trend in Hubei province during2004-2011. Spatial autocorrelation analysis results indicated that there was a certain spatial cluster in malaria case in Hubei province. High-risk areas still exist. The results showed that the high-risk areas of malaria prevalence in Hubei province were mainly located in A. sinensis and A. anthropophagus distribution areas.(2) The incidence of malaria in the whole province and high-incidence counties achieved the better effect by fitting using the meteorological factors. The constructed discriminant function could accurately determine the occurrence time of high, medium and low malaria incidence month in Zaoyan County.(3) The fitting effect of the constructed ARIMA model on the incidence of malaria was satisfactory in Hubei Province, the prediction effect was good, the model to predict the incidence trend of local malaria in Hubei province.