Study On Epidemiological Characteristics,Meterological Effects And Predictive Models Of Hand,Foot And Mouth Disease In Different Regions

BackgroundHand,foot and mouth disease(HFMD)is a contagious disease caused by enterovirus infection.Between 2009 and 2018,1.5 to 2 million cases of the disease were reported annually in China,resulting in a huge disease burden.HFMD has obvious seasonality,but the seasonality varies in different regions.Previous studies have shown that meteorological factors may be one of the factors driving seasonal fluctuations of HFMD.In recent years,the changes of meteorological and other environmental factors have brought about serious impacts on human production and life,and directly or indirectly affected the epidemic pattern,frequency,and intensity of infectious diseases,which seriously endangers human survival and health.Therefore,it is of great public health significance to consider the influence of meteorological factors when formulating prevention and control policies of HFMD.Existing studies on the relationship between meteorological factors and HFMD are mainly limited to some regions or use single statistical analysis method,which is easy to ignore the regional differences in the influence of meteorological factors on HFMD and the interaction effects of meteorological factors.Based on the effects of meteorological factors,it is still an open question how to further analyze the driving effect of meteorological factors on the seasonality of HFMD based on the transmission mechanism of the disease.At present,some studies have explored the transmission dynamics of HFMD based on SIR model or its extended model,but few studies have fully considered the influence of meteorological factors,so it is difficult to get the research results closer to the real scene.The predictive model of HFMD based on the influence of meteorological factors can predict the epidemic dynamics of HFMD timely and accurately,which is conducive to the development of disease prevention and control measures in advance and improve prevention and control efficiency.The predictive models of HFMD constructed by existing studies are mainly used to verify the predictive effect in limited regions or seasonal situations,and it is difficult to guarantee the stability in other regions or seasonal situations.Therefore,based on the reported incidence data of HFMD in Hangzhou,Xi’an and Zibo from 2010 to 2019,this study systematically and comprehensively evaluated the influence of meteorological factors on the seasonal prevalence of HFMD based on the seasonal epidemic characteristics of HFMD,and built a dynamic model to further explore the seasonal driving effect of meteorological factors.Based on meteorological factors,a predictive model with good performance and stability is constructed.Through this study,the seasonal epidemic regularity of HFMD can be revealed,the driving effect of meteorological factors on seasonal HFMD can be quantitatively analyzed,and the best predictive model for predicting the epidemic dynamics of HFMD can be determined,which is helpful for local governments and relevant prevention and control departments to scientifically evaluate and predict the seasonal epidemic dynamics of HFMD based on meteorological factors,and formulate corresponding prevention and control measures in advance to reduce the prevalence and spread of disease.Data and Methods1.Data collectionData on reported cases of HFMD in Hangzhou,Xi’an and Zibo City from 2010 to 2019 were obtained from the China Disease Prevention and Control Information System.Meteorological data for the same period were obtained from European Centre for MediumRange Weather Forecasts.Demographic and socioeconomic data for the same period were obtained from municipal or county-level statistical yearbooks.All data were quality controlled by relevant departments and professionals,and the above data were organized into datasets of different temporal and spatial scales according to the purpose of the study.2.Statistical analysis methods(1)To analyze the epidemiological characteristics of HFMD.The prevalence and distribution of reported HFMD cases in three cities were described;spatial autocorrelation analysis and spatio-temporal scanning analysis were used to explore the spatial distribution characteristics and spatio-temporal aggregation of the disease.(2)To investigate the influence of meteorological factors on the seasonal incidence of HFMD.Firstly,the correlations between the incidence of HFMD and meteorological factors were established to determine the included meteorological factors.Then,a two-stage model was constructed to analyze the effects of temperature and relative humidity on the incidence of HFMD at the county level,city level and overall level,identify potential effect modifier factors,identify sensitive population through subgroup analysis,and conduct sensitivity analysis.Finally,the binary surface of temperature and humidity effects was obtained by the generalized additive model to identify the interaction of meteorological factors on the incidence of HFMD,and the comprehensive effects of temperature and relative humidity were explored by comprehensive indexes.(3)The seasonal driving effect of meteorological factors on HFMD was investigated based on dynamic model.Firstly,based on the natural history of HFMD,The SEIAR infectious disease dynamics model was constructed,which divided the population into five categories:susceptible(S),exposed(E),symptomatic infected(I),asymptomatic infected(A)and recovered(R).Meteorological factors and school holidays were incorporated into the model through seasonal functions under different situations.Secondly,simulation data were used to verify the validity of the model and the accuracy of parameter estimation.Finally,the best model was determined by comparing the fitting effects of models in different scenarios,so as to analyze the driving effect of meteorological factors and evaluate the prevention and control effects of different intervention measures based on the simulation of the best model.(4)To construct and compare the seasonal predictive models of HFMD.Based on the seasonal epidemiological characteristics of HFMD,nine models were constructed to predict the epidemiological dynamics of HFMD,including seven single models(RF,GBM,XGBoost,SVM,ARIMA,LSTM,and LR),one two-level ensemble model(SL),and one three-level ensemble model(DEML).The accuracy,stability and running time of these 9 models in predicting the epidemic dynamics of HFMD in the next 1～4 weeks under different scenarios were comparatively analyzed by simulated data in 18 scenarios and actual data in 3 representative cities to finally determine the best predictive model.Results1.The average annual reported incidence rates in Hangzhou,Xi’an and Zibo Cities during the study period were 110.6,250.6 and 124.5 per 100 thousand,respectively,with reported severity rates of 0.5‰,3.1‰ and 0.8‰,respectively.Temporal distribution showed that the annual peak of cases in Hangzhou was from June to July,with irregular secondary peaks in the second half of some years;Xi’an had two annual peaks of cases,with the main peak in May to June and the secondary peak in October to November;Zibo had only one significant peak of cases in June to July every year.The population distribution showed that cases were mainly concentrated in the population under 6 years old,with the highest proportion of cases aged 1～3 years was 67.4%,70.1%and 64.2%in three cities,respectively.The sex ratio was roughly 1.50:1,and the children were scattered more than those in nursery care.The spatial distribution results showed that the incidence rate was higher in rural-urban fringe of Hangzhou and Xi’an than in the urban areas,but the incidence rate was higher in the central areas in the surrounding counties of all three cities.There were significant spatial autocorrelation and temporal aggregation in all three cities.2.The cumulative effect of temperature on the incidence of HFMD was roughly in the shape of an inverted "V",and the strongest effect was observed at 28.2℃,when the relative risk was 1.60(95%CI:1.30～1.96).The effect of relative humidity was roughly in the shape of"U",and the effect was strongest at 88.9%,with a relative risk of 1.58(95%CI:1.47～1.70).The effect of temperature varied across counties,with GDP per capita,gender ratio,geographic location,mean precipitation,and mean wind speed as identified effect modifiers.diaspora children aged 0 to 2 years were more sensitive to changes in temperature.Among the comprehensive indicators considered,the humidity index was more comprehensive in reflecting the effects of temperature and humidity.3.Seasonal prevalence of HFMD is promoted by a combination of seasonal changes in the infectiousness of infected individuals and pathogens in the environment.0.71(95%CI:0.44%～0.93)of infected individuals under 6 years of age were reported,and 86.87%(95%CI:86.34%～87.37%)of reported cases were due to pathogens in the environment.Meteorological factors affect the seasonal epidemics of diseases by simultaneously driving the two seasonal promotion factors described above,with meteorological factors and holidays accounting for 39.71%of the seasonal variation of diseases.When the vaccination rate was 70%,the infectiousness of pathogens in the environment was reduced by 70%,and the infectiousness of infected individuals was reduced by 70%,the total number of cases was reduced by 71%,70%,and 8%,respectively.4.In predicting the epidemic dynamics of HFMD in the next 1～4 weeks,the ensemble models(SL and DEML)had the top 3 prediction performances in both simulated and actual data.The accuracy of DEML prediction was slightly better than SL,followed by GBM,XGBoost,and SVM,while the prediction performance of RF,ARIMA,and LSTM varied widely across different scenarios,and the prediction performance of LR was the worst.The SL ran 40 minutes longer than the baseline model combined,and the average running time of DEML is 1.65 times that of SL.Conclusions1.HFMD had alternating peaks,double peaks and single peaks in Hangzhou,Xi ’an and Zibo,respectively.Children aged 1 to 3 years,males,and children living in the diaspora predominate among the reported cases in all 3 cities.The incidence rate was higher in ruralurban fringe than in urban areas,and the incidence rate was higher in the central areas of all surrounding counties.2.The relationship between temperature and relative humidity and the incidence of HFMD was inverted V shape and U shape,respectively,but the effect varies among different regions.GDP per capita,sex ratio,geographic location,average precipitation and average wind speed are modifiers of the effect;children aged 0-2 years are more sensitive to meteorological changes;the humidity index can comprehensively reflect the effect of temperature and humidity.3.Seasonal epidemics of HFMD are promoted by seasonal changes in the infectiousness of infected individuals and pathogens in the environment,and meteorological factors influence seasonal epidemics of the disease by driving both of these seasonal promoters.The epidemic intensity of HFMD can be effectively reduced by vaccination and by reducing the infectiousness of pathogens in the environment.4.SL has stable and better predictive performance in different scenarios,and it can predict the epidemic dynamics of HFMD in the next 1～4 weeks in a timely and accurate manner with a relatively short running time,which is the best model for predicting the epidemic dynamics of HFMD identified in this study.Policy suggestions1.In the prevention and control of HFMD,attention should be paid to 1～3 years old,male and scattered children.The focus of prevention and control in large cities such as Hangzhou and Xi ’an can be tilted to the rural-urban fringe,while the focus in cities such as Zibo should be placed in the political centers of counties and districts.2.Relevant departments of prevention and control of HFMD should always pay attention to changes in meteorological factors(especially temperature and humidity)and strengthen prevention and control efforts under high temperature and humidity environment,especially for the protection of children aged 0～2 years,so as to reduce the prevalence of HFMD caused by environmental changes.3.Pathogens that persist in the environment play important roles in the epidemic of HFMD.In order to reduce the infectivity of pathogens in the environment and effectively reduce the epidemic intensity of diseases,kindergartens,children’s activity centers and other children’s activity areas as well as toys,towels and other items easily accessible to children should be disinfected.4.In the seasonal epidemic prediction of HFMD,the predictive model with two-layer structure(Super Learner)has stable and good prediction performance under different situations.After increasing the number of layers of the model framework,the prediction accuracy was little improved,but the calculation time was greatly improved.Therefore,Super Learner can be used as an important tool for dynamic prediction of HFMD epidemic under the comprehensive consideration of prediction accuracy,stability,and calculation time.Innovativeness1.Based on the natural history and seasonal epidemic characteristics of HFMD,a dynamic model integrating meteorological driving factors was constructed,and the driving effect of meteorological factors was determined through simulation experiments.The influence of meteorological factors on seasonal epidemic and transmission of HFMD was first explored from the perspective of theoretical epidemiology.2.Nine predictive models of HFMD with single-layer,double-layer and three-layer structures were constructed.The stability and practicability of the predictive models were compared through simulation and actual data,so as to determine the optimal modeling strategy for predicting HFMD,which also provides reference for the selection of predictive models of other infectious diseases.3.Descriptive,analytical,and theoretical research methods were used to systematically and comprehensively explore the influence of meteorological factors on the seasonal prevalence and transmission of HFMD.It established a research system for precise prevention and control of HFMD based on meteorological factors,providing scientific basis for prevention and control of HFMD.