Spatial And Temporal Characteristics And Prediction Of Hemorrhagic Fever With Renal Syndrome In Three Northeastern Provinces Of China

Objective: To analyze the epidemiological characteristic and spatial-temporal distribution of Hemorrhagic fever with renal syndrome(HFRS)in three northeastern provinces of China,spatial-temporal statistical methods and prediction models were uesd to explore the relationship between the HFRS and meteorological factors in northeastern China,and the study trained a machine learning prediction model for HFRS,to evaluate its effiency and provide a reference basis for the prevention and control of HFRS in China.Methods:(1)The data of hemorrhagic fever with renal syndrome in Heilongjiang,Jilin and Liaoning provinces in northeast China from 2006 to 2020 were collected from the Chinese Infectious Disease Prevention and Control Information System,and the monthly meteorological data of the same period were collected through China Meteorological Administration.(2)To study the global autocorrelation and local autocorrelation of hemorrhagic fever incidence in China by using spatial autocorrelation method,and to detect "high value-high value" or "low value-low value" clusters of HFRS incidence.(3)To analyze the relationship and interaction between HFRS and meteorological factors in northeastern China by using geodetector model;(4)To select the monthly incidence data of HFRS in Jiamusi City from 2006 to 2020 and fit it to establish prediction models.(4)The monthly data of HFRS in Jiamusi city from 2006 to 2020 were selected and fitted to establish vector machine regression model(SVR)and SRAIMA model,including data pre-processing,model training and validation,parameter adjustment and model performance evaluation,etc.;(5)The nine meteorological factors related to HFRS were screened by Spearman correlation analysis to construct SVR model,and the validation set of HFRS data was verified by root mean squared error(RMSE).The root mean squared error(RMSE),mean absolute percentage error(MAPE)and mean absolute error(MAE)were used to evaluate the model performance and analyze the prediction effects of the two models.Results:(1)The number of reported HFRS cases in northeast China from 2006 to2020 was 52,655,with a decreasing number of reported cases year by year and a summer-winter peak,and most patients with HFRS were aged between 30-59 years.,with a male to female incidence ratio of about 3.4:1.(2)The spatial global autocorrelation of HFRS was explored by Moran’s I index,with a significant spatial clustering(Z>1.96,P<0.05).local autocorrelation analysis using Moran’s I index detected five clusters of "high value-high value",mainly in Jiamusi City,Shuangyashan City and Jixi City.A total of 8 "low value-low value" clusters were detected,mainly in Baicheng and Daqing.Overall,the distribution of HFRS in this region was higher in the east than in the west.(3)The geodetector analysis showed that the ground temperature(4.18%),precipitation(3.54%)and air temperature(3.31%)were the top three most important meteorological variables affecting HFRS in the northeastern region,where there was a significant enhanced interaction effect(10.32%)between the ground temperature in the following month and the air pressure in the lagging four months.(4)From 2006 to 2020,a total of 4,525 cases of HFRS were reported in Jiamusi.wind speed had a negative correlation(r =-0.72)with HFRS and HFRS has the negative correlation(r =-0.51)with precipitation.While positive correlations were observed with pressure,maximum evaporation,and minimum evaporation,with minimum evaporation having the strongest positive correlation(r = 0.71)with HFRS,followed by maximum evaporation(r = 0.55),and the correlation between pressure and HFRS is 0.55.(5)The time series model SARIMA(2,1,5)(1,1,1)12 had a predictive efficacy with RMSE = 33.74 and MAPE = 7.76.And the results of the SVR prediction model were consistent with the actual trend,the total number of fitted HFRS cases 3,809,and the model efficacy showed that the overall RMSE = 19.86 and MAPE = 6.73,showing better predictive efficacy compared to the SARIMA model.Conclusion: In this study,we analyzed the dynamics and characteristics of HFRS in northeast China from 2006 to 2020 using spatio-temporal statistical analysis methods,geographic probes and support vector machine regression and SRIMA prediction models.The analysis found that:(1)The HFRS epidemic in northeast China is periodical and seasonal,and cases are mainly concentrated in summer and winter.(2)The HFRS is spatially aggregated,and the "high value-high value" aggregation area is mainly distributed in the eastern cities of northeastern China.(3)In the analysis of meteorological risk factors,we found that ground temperature,air temperature and precipitation are the most influential factors for the overall HFRS,and there is a lag effect which varied by region.(4)The meteorological factors were incorporated to establish two prediction models for Jiamusi city,and the results showed that the SVR model had higher prediction accuracy compared with the SARIMA model,which is beneficial to provide analytical ideas and guidance basis for HFRS surveillance and early warning and prevention and control decisions.