Study On Influencing Factors And Prediction Models Of HFRS In Different Climate Zones Of China

BackgroundHaemorrhagic fever with renal syndrome(HFRS)is a kind of zoonosis disease caused by different kinds of hantavirus.Transmission of the virus to humans is mainly through inhalation or exposure to the excreta(urine,feces or saliva)of infected rodents.China is the country with the largest number of HFRS cases in the world.About 70%of HFRS cases in China were caused by Hantaan virus(HTNV)and Seoul virus(SEOV),among which,some developed into moderate or severe cases,causing serious sequelae or even death.The average fatality rate of HFRS is 1%-15%around the world.Since the 80s of the 20th century,the incidence of HFRS in China has shown a more obvious seasonal distribution change,which was characterized with the type of Apodemus epidemic areas that that mainly occurred in autumn and winter to mixed epidemic areas with two peaks in spring,autumn and winter.The change of epidemic area type is closely related to the genotype of virus,the species of dominant host animals,natural geographical environment,human production activities and behaviors.However,with the increasing number of HFRS cases,the expansion of epidemic area and change of epidemic area types,we need to update our knowledge of the epidemic characteristics HFRS,and also understand current temporal and spatial distribution characteristics of HFRS.As a vector borne infectious disease,HFRS is affected by many factors:environmental factors,rodent population(vectors),interactions between human and animal hosts,and dynamics of hantavirus.Among them,meteorological factors play a very important role in the spread of HFRS.Meteorological factors the prevalence of HFRS directly or indirectly.Epidemiological studies showed obvious seasonal regularity of HFRS,which also suggests that the HFRS is climate sensitive.Meteorological factors have important impacts on the survival,reproduction,distribution and population change of rodents.With the background of global warming,the ideal breeding area of vectors may be expanded,and climate warming may be conducive to the extension of the breeding period of host animals.Human activities is restricted by weather conditions and seasonal changes,thus affecting the contact opportunities between human beings and vectors.Environment factors such as geographical environment and ecology also have an impact on the epidemic of natural focus diseases.Tree cutting,land acquisition,road construction and bridge construction behind the urbanization construction will directly affect the population numbers and distribution of host animals,which may increase the density of local vectors and cause disease outbreaks.Nowadays,with the development of social economy,the floating population is increasing.The development of logistics has an important impact on the spread of infectious diseases.Although there have been a lot of quantitative analysis on the relationship between meteorological factors and HFRS,most of these studies are based on some areas of a city/Province,and there are few studies on the interaction between HFRS meteorological factors.Therefore,this study focuses on the impact of meteorological factors on HFRS,as well as the interaction and marginal effect between meteorological factors.With the increasing demand of people's health,disease prediction has been developed as an important aspect of disease prevention and control.The influencing factors of infectious diseases are numerous and the relationship between them is complex and the accuracy of model prediction is an extremely important aspect of disease prediction.Machine learning as a new analytical method has been developed rapidly and widely used with the development of information technology and the advent of the era of big data.How to use big data to predict and early warn the epidemic situation of infectious diseases has become a research hotspot in the field of disease prevention and control.In order to explore the practicability of machine learning in the prediction of HFRS and provide a new idea for the prediction of HFRS,this study takes HFRS cases in various climate zones of China as the research object,constructs a random forest regression prediction model,and compares it with the traditional prediction model to evaluate the fitting and prediction effect in each climate zone.Objectives1.To analyse the epidemic characteristics of temporal and spatial distribution of HFRS cases in China from 2006 to 2016.2.The influencing factors of HFRS prevalence in different climate areas are discussed by integrating meteorological factors,social factors and geographical environment related indicators.3.To construct prediction models of HFRS with random forest regression and provide analysis method reference for further accurate prediction of HFRS occurrence in the future.MethodsSurveillance data of HFRS cases between 1 January 2006 and 31 December 2016 were provided by Chinese centre for disease control and prevention.Daily meteorological data from 839 climate monitor stations during 2006-2016 across China were obtained from the China Meteorological Data Sharing Service System.NDVI grid data from 2006-2016 and altitude grid data were connected with the digital map of China at prefecture level to establish the geographic information database of HFRS epidemic.Temporal and spatial distribution of HFRS in China was studied by spatial analysis.The influencing factors of HFRS were analysed by generalized estimation equation and the incidence of HFRS was predicted by random forest regression model.The software used in the research includes ArcGIS10.2,SaTScan 9.1,Stata 16.0 and R 3.4.3.Results1.There were totally 121,494 cases reported in China during 2006-2016 with the average annual incidence rate of 0.89/10 million.The incidence of HFRS in China has declined slightly,the annual incidence has dropped from 1.16/10 10000 in 2006 to 0.64/10 million in 2016,but the 2012-2013 year has seen a relatively obvious short-term increase.2.There were more male cases than females,and the sex ratio of male and female cases is 3:1.From the perspective of occupational distribution,the cases are mainly farmers.The age of onset was mainly between 20-40 and 41-60 years old group,accounting for 33.08%and 46.04%of the cases respectively.The death cases between 41-60 years old accounted for 53.83%of the total death cases.3.The time intervals between the onset and the diagnosis of HFRS cases in different provinces during 2006 to 2016 were analysed.Results showed that the average interval from the onset to the diagnosis of HFRS cases in China was 7.6 days.In Heilongjiang,Jilin,Hebei and Shaanxi provinces,the interval between the onset and the diagnosis of HFRS cases is the shortest with an average of 5 days.The interval was 6 days in Liaoning and Shandong Province,8 days in Zhejiang,Jiangxi and Hubei Province,10 days in Guangdong Province,and 9 days in other provinces.4.The distribution of HFRS cases in China showed spatial correlations.From January 2006 to December 2016,the incidence of HFRS at Prefecture level was scanned and the results showed that there were 13 aggregation areas.Although the hot spots of HFRS differed by regions and time,they were mainly distributed in Heilongjiang province,Jilin province,Liaoning province,Shandong Province,Shaanxi Province,Zhejiang Province,Jiangxi Province and Hubei Province.5.HFRS showed obvious seasonal features and the variation differed by regions.In the temperate zone,relatively more cases were observed in autumn,while for the warm temperate and subtropical zones HFRS peaked in winter and spring.During 2006-2009,the number of HFRS cases in the temperate zone was larger than that in the warm temperate zone and subtropical zone.Since 2009,the cases of HFRS in the warm temperate zone has increased significantly and exceeded that in the temperate zone to become the climate zone with the most cases.the incidence of HFRS in the middle temperate zone was higher than that in the warm temperate zone and subtropical zone.From 2006 to 2016,the number of cases in temperate zone decreased slowly.6.The maximum coefficients of the association between monthly mean temperature and counts of HFRS from GEE model in different climate zones of China were calculated after controlling for precipitation,relative humidity,seasonality and long-term trends.The maximum lagged effects of temperature in temperate zone,warm temperate zone and subtropical zone was 1-month,2-month,3-month,respectively.7.There was interaction between mean temperature and precipitation for temperate zone after adjusting relative humidity,seasonality and long-term trends at a 1-month lag.And interaction between mean temperature and relative humidity at a 2-month lag was observed in warm temperate zone.8.Results from GEE model showed the relationships between HFRS incidence and temperature,relative humidity,altitude,per capita cultivated land area,GDP and HFRS incidence were statistically significant(p<0.05)in temperate zone.Among which,temperature,relative humidity,altitude,GDP were protective factors,and per capita cultivated land area was risk factors.For warm temperate zone,the relationships between HFRS incidence and temperature,relative humidity and altitude were statistically significant(p<0.05),and temperature and altitude were protective factors,while relative humidity was risk factors.In subtropical zone,the factors that showed statistical significance for HFRS are precipitation,altitude and GDP,among which precipitation was risk factors,altitude and GDP were protective factors.9.The prediction of the random forest regression model constructed by employing NDVI and meteorological indexes(temperature,humidity and precipitation)in each climate zone were similar to each other,and the fitting effect between each climate zone showed better from north climate zones to south.The incidence rate was highly estimated by random forest regression model in the temperate zone,while closer to the true value for the warm temperate and subtropical zones.The estimated incidence rate was generally lower obtained by generalized equation model for all the focused climate zones.Conclusions1.The incidence of HFRS in China has been decreasing in the past 2006-2016 years.The cases were mainly male and farmers.And most cases aged 20-60 years old.Cases were reported in 31 provinces and autonomous regions in China.More cases occurred in the north,especially in northeast provinces(Heilongjiang,Jilin,Liaoning province),Hebei Province,Shaanxi Province,Shandong Province.In the south,HFRS cases were distributed mainly in Zhejiang Province,Jiangxi province and Hunan Province.The spatial distribution of HFRS cases showed significant correlation,and spatial-temporal clustering areas of HFRS cases were detected in the above provinces.2.The time intervals between the onset and the diagnosis of HFRS cases were shorter in the provinces with high incidence of HFRS in the north of China.And the time interval between the onset and the diagnosis of HFRS cases in the south China is generally longer than that in the north,which is positively related to the number of confirmed cases.3.The trend of HFRS incidence in different climate zones has changed.During 2006-2009,the number of HFRS cases in the temperate zone was larger than that in the warm temperate zone and subtropical zone.Since 2009,the cases of HFRS in the warm temperate zone has increased significantly and exceeded that in the temperate zone to become the climate zone with the most cases.It suggested to pay attention to strengthen the prevention and control of HFRS in the warm temperate zone.4.In the temperate zone and warm temperate zone,the relationship between temperature and HFRS was nonlinear,while in the subtropical zone,the relationship between temperature and HFRS was linear.From the north to the south of China,the optimal lag period of temperature in these three climate zones was 1-month,2-months and 3-months respectively.The determination of the optimal lag period can provide clues for the early warning of HFRS.5.The influence of meteorological factors on HFRS in the temperate zone showed that there were interactions between temperature and precipitation.Cold temperature and precipitation may increase the incidence of HFRS in the temperate zone.Interactions effect between temperature and relative humidity were also found in warm temperate zone,high temperature and high relative humidity may increase the incidence of HFRS in the warm temperate zone.6.The results showed that GDP in the middle temperate and subtropical regions was the protective factor of HFRS.There was a negative correlation between HFRS and altitude in the temperate zone,warm temperate zone and subtropical zone,which means the risk of HFRS is higher in the low altitude area.A positive correlation between the per capita cultivated land area and HFRS was observed in temperate zone,thereby the risk of HFRS is higher in the area with large per capita cultivated land area.7.The prediction accuracy of random forest regression model is better than that of generalized estimation equation.NDVI can be used to predict HFRS instead of meteorological indexes(temperature,humidity and precipitation).Significance and innovation1.The design,analysis,comparison and model construction of this study were based on different climate zones in China.2.The results of this study showed that the best lag period of temperature is one month,two months and three months respectively in the temperate zone,warm temperate zone and subtropical zone from the north to the south of China.There was interaction between temperature and precipitation in the middle temperate zone,and between temperature and relative humidity in the warm temperate zone.3.In this study,the prediction model of HFRS was constructed by using the NDVI and the main meteorological indexes(temperature,humidity and precipitation).It was firstly founded that the prediction model constructed by using the NDVI is similar to the meteorological model constructed by using the above meteorological indexes in each climatic zone.The fitting effect between each climate zone was better from north to south.