Study On The Spatiotemporal Distribution Characteristics And Prediction Models Of Acute Hemorrhagic Conjunctivitis In Chongqing From 2004 To 2018

Objectives:This study based the acute hemorrhagic conjunctivitis（AHC）monitoring data in Chongqing from 2004 to 2018.We analyzed the epidemic characteristics and population distribution of AHC to understand the epidemic variation and change rules.At the same time,spatial and temporal analysis was used to understand the geographical distribution and spatial aggregation of AHC.Based on the monitoring data of AHC in Chongqing,the Seasonal Autoregressive Integrated Moving Average（SARIMA）and the Wavelet Neural Network（WNN）models were constructed to predict the incidence of the disease in a short term,and the prediction effects of the two models were evaluated,so as to provide reference for the relevant departments to formulate scientific prevention and control measures.Methods:1.Collected and sorted out the data,studied the distribution characteristics of AHC cases in Chongqing during 2004-2018 in different regions and different populations at different times through descriptive statistical analysis,and charts were drawn by Excel 2007.2.Geographic information system（GIS）and spatial analysis technology were used to analyze the spatial distribution characteristics and development trend of AHC in Chongqing.The total population and incidence of AHC in different districts and counties of Chongqing in different years were used to calculate the incidence in different regions from 2004 to 2018.The incidence information of AHC in Chongqing was correlated with geographical information by using ArcGIS 10.4 software,and the map was drawn and the spatial analysis data was visualized.Different prevalence degrees of AHC were represented by different colors,so as to intuitively judge the prevalence differences of AHC in different regions.Spatial Autocorrelation Analysis was performed by GeoDa 1.1software to determine the existence of clustering or other Spatial characteristics and to reflect whether regional units are high or low clustering.SaTScan 9.4 software was used to perform a scan statistic to analyze whether there was an aggregation tendency or trend in the spatial and temporal distribution of AHC in Chongqing,andα=0.05.3.SARIMA prediction model was established by R 3.5.1 software to fit the monthly incidence of AHC in Chongqing from 2004 to 2017,and to predict the trend of monthly incidence of AHC from 2018 to 2020.4.Three WNN prediction models were established by Matlab 2012a software to select the optimal model and predict the monthly incidence of AHC in Chongqing in 2018.Discussions:1.Epidemic characteristics of Chongqing AHC from 2004 to2018A total of 30,686 cases of AHC were reported in Chongqing from2004 to 2018,with an annual reported incidence of 7.04/100,000.The incidence was relatively stable from 2004 to 2006,2007 presented a small peak（10.75/100,000）.The incidence dropped in 2008 and 2009.A significant peak（41.17/100,000）in 2010 and presenting an obvious outbreak trend.With the exception in 2014（8.36/100,000）,the incidence was basically flat from 2011 to 2018,and no deaths were reported in 15years.The average sex ratio of male-to-female in 12,565 cases was 1.44:1,higher than that of the total population of Chongqing in 2018.In terms of age composition,patients aged 10-20 years were the most（31.34%）and elderly patients over 75 years were the least（1.59%）.In terms of population classification,the top five cases of AHC were students（39.75%）,farmers（28.05%）,kindergarten children（6.40%）,scattered children（5.09%）and domestic workers and unemployed（4.91%）.The onset time of AHC has obvious seasonal characteristics,with the peak in September.2.Spatial and temporal distribution characteristics of AHCExcept for 2012（Moran’s I=0.2442,P=0.015）and 2013（Moran’s I=0.1661,P=0.039）,the global spatial autocorrelation of AHC incidence in other years did not exhibit significant global correlation.It means that the distribution of AHC in all regions is not spatially dependent in most years.Locally autocorrelated high-high aggregation areas only existed in Chengkou and Wuxi in 2011 and 2012,and Chengkou in 2013 and 2015.The epidemic situation in the suburbs was higher than that in the main city,and no hot spots were found in other years.From 2004 to 2018,four aggregation areas were detected respectively,and the aggregation time was September 1,2010 to September 30,2010.The most likely cluster were located in Chengkou,Wanzhou,Yunyang,Kaixian,Liangping and Wuxi.The second cluster areas were located in Yubei,Dadukou,Beibei,Nanan,Yuzhong,Changshou and Shapingba.The 2nd secondary cluster were Rongchang,Dazu,Yongchuan,Tongliang,Bishan and Tongnan.The 3rd secondary cluster were located in Qianjiang,Pengshui,Youyang,Shizhu,Wulong,Fengdu and Xiushan.3.Short-term prediction of AHC trends in Chongqing based on ARIMA modelThe SARIMA（1,1,2）（0,1,1）₁₂2 model was fitted based on the monthly incidence of AHC in Chongqing from 2004 to 2017,and the actual incidence and the incidence of the fitted value in 2018 were0.42/100,000 and 0.47/100,000,respectively.According to the predicted results,the monthly incidence of AHC in Chongqing will increase slightly in the next two years,and the incidence of AHC in 2019 and2020 are predicted to be 0.48/100,000 and 0.51/100,000.4.Prediction of monthly incidence of AHC in Chongqing in2018 based on the WNN modelAccording to various evaluation indexes,the 5-8-1 network structure was selected as the optimal model,and the monthly incidence of AHC in2018 was predicted.The mean square error（MSE）,root mean square error（RMSE）and mean absolute percentage error（MAPE）were 0.0282,0.0102 and 2.26%,respectively.The model had a good prediction effect and was better than the SARIMA（1,1,2）（0,1,1）₁₂2 model.Conclusions:The study found that except for the high incidence in 2007 and 2014and the outbreak in 2010,the incidence of AHC in Chongqing was generally stable in the past 15 years.Individuals aged 10-19 years,males,students and farmers were more susceptible to AHC,with the incidence peak in September.The northeast surrounding areas and the main urban areas should be the priority areas for prevention and control.The SARIMA prediction model indicates that the incidence of AHC in Chongqing will continue to increase slightly from 2019 to 2020.The WNN model has a better prediction effect on the incidence rate of AHC in Chongqing than that of SARIMA model.Relevant departments should strengthen health management and monitoring of key groups and hot spots,to promote the rational allocation of medical resources in various regions,and using targeted prevention and treatment measures in advance according to the prediction tips,so as to avoid causing large-scale epidemic outbreaks of infectious diseases.